Get started
AI plan
The latest read of your Meta campaigns.
AI plan · Google Ads
The latest read of your Google Ads campaigns.
Installs by app
Paid, organic and undetermined.
Top campaigns
Where your ad money went in this period.
Spend against proceeds
What went out on ads and everything the stores earned you in the same days.
By period
The figures behind the chart. Paid, organic and undetermined add up to all proceeds.
By app
The whole period, for each app.
How to read these numbers
Total ROAS is all proceeds over ad spend. It counts every payment in these days: organic users, renewals, and users who arrived before the period. It says whether the stores cover what you spend on ads, not whether the ads themselves pay back.Paid ROAS counts only payments from users whose install is credited to an ad, by the same rules as Install sources. On iOS an ad install Bridge could not match counts as organic, so read Paid ROAS as the low end and Total ROAS as the high end: the real return from ads is between them. When a fifth or more of the proceeds come from users Bridge has no install record for (installed before the app shipped a build with the Bridge SDK), the card says it is a floor. When Google Ads is all the spend, the card also gives Google’s own figure: its conversions’ purchase value over its cost, which reads high.
Both are cash in the period, so an annual plan counts in full on the day it is paid, and users from this period’s ads keep paying later. To judge a campaign on the users a window brought in, see .
Left after ads is proceeds minus ad spend. It is not profit: no other cost is taken off.
Campaigns
Cash in this period: what users from each campaign paid in these days, against what it spent in them. To judge a campaign on the users a window brought in, see .
Record ad spend to see ROAS
Revenue and ROAS are shown per currency, never summed across them: a
total mixing EUR and CLP is a number in no currency at all. ROAS appears
only where spend exists in the same currency as the revenue.
Spend is the one figure this system cannot observe. It lives in Meta's
Marketing API, which needs an ads_read token on the ad
account — a different credential from the dataset token that sends
conversions. Until that is connected, enter it below and ROAS becomes
real; the numbers above it are measured either way.
Spend for one campaign on one day. Saving the same day again replaces it.
Scheduled ads
Seasonal ads added through the Bridge connector with dates: Bridge switches each on on its first day and pauses it after its last. Below them, what it did in the last two weeks.
Funnel by ad
Every campaign, from click to purchase. Open a campaign to see its ads.
Price tests
New users split between the current offering and another of yours at other prices, each put on theirs in RevenueCat, and judged on what they pay in their first 30 days. Nothing is made in the App Store or Google Play. How it works.
Loading…
Offerings
What RevenueCat serves: the current offering is what a new user sees, unless a test or an override gives them another.
Loading…
Products
Each store’s subscriptions, with the price in the United States, the store’s state and how many people pay for each now. A price test reprices one nobody holds before it makes a new one.
How the products are named
Read off this app’s own product ids, per store. A new product for a test is named the same way.
Store listing
Each language’s text as buyers see it, and on the App Store as the version being prepared has it. Edit as many languages as you like, then preview: nothing reaches a store until you send it.
Loading…
Screenshots
Each device’s screenshots in the language picked above, replaced in the order you give: images from this computer, Figma frames (exported at the size the store takes) or links.
Custom product pages
Other App Store pages for an ad, a season or an audience, with their own screenshots and promotional text. A page made here gets its own ad link under Setup.
Changes
What was changed through Bridge, here or by Claude, with the text before: put it back in one step.
Check a keyword
How popular it is on the App Store (Apple’s score, 0 to 100, relative, not a count of searches), whether your Apple Ads show for it, and whether one of their blocked searches would stop it; with Apple’s related suggestions.
Keywords you target
This app’s Apple Ads campaigns, each with its ad groups’ keywords and the searches it blocks. For each keyword: how popular it is on the App Store, its bid, and over the last 30 days Apple’s spend, taps and installs beside the trials and proceeds of the users Apple credits it with.
Searches that showed your ads
What people typed on the App Store when your ads showed, most spent first, and the keyword each matched (or Search Match). A search that is not a keyword yet can be added to its ad group, or blocked.
Check a keyword
How often people search for it, whether your Search campaigns show for it, and whether one of their blocked searches would stop it; with Google’s related ideas.
Keywords you target
Every keyword of this app’s Google Search campaigns over the last 30 days: Google’s spend, clicks and conversions, beside the installs, trials and proceeds of the people its clicks brought, as Bridge saw them. Searches a month and the top-of-page bid are Google’s Keyword Planner figures.
Searches that showed your ads
What people typed when your ads showed, most spent first, and the keyword each matched. A search that is not a keyword yet can be added as one, or blocked.
Blocked searches
Negative keywords: searches your ads stay out of, for the whole campaign or one ad group.
Install sources
Each app's installs split into paid, organic and undetermined. Click an app or a group to fold it.
What paid, organic and undetermined mean
Paid means something names an ad network: Bridge matched the install to one of its ad clicks (named by where the ad ran), Google Play's install referrer carries a Google Ads click id or Meta's own referrer, or RevenueCat reports Apple Search Ads. Facebook app's own referrer is a Meta ad install where the Facebook or Instagram app opened Google Play itself instead of going through the Bridge link, so the campaign behind it is not known.Organic means nothing named a paid source, which is also how commercial MMPs count it. On Android, Google Play usually says how the user arrived. On iOS, Apple gives apps no install referrer, so an iOS install from an ad that Bridge could not match is counted here too: read iOS organic as an upper bound.
Undetermined installs are ones Bridge saw but could not decide: two ad clicks too close to call, or a phone behind a VPN or iCloud Private Relay. They are left out of both rather than guessed.
Apple Search Ads shows only for users who started a trial or purchased, because that is when Apple's attribution reaches Bridge through RevenueCat.
Revenue and Proceeds count payments by the date they were earned, the same way the Campaigns table does, and credit each to the source of the user's install: the payment's own ad click when it has one, otherwise the first install Bridge saw for that user. Payments from users Bridge has no install for, most often subscribers from before the app shipped Bridge, are Undetermined. Both are in USD; Proceeds is what is left after the store's commission and taxes, using RevenueCat's own figure.
Payouts
Each store’s payout periods, the day it pays and about how much: proceeds in USD after the store’s cut and tax, refunds taken off. A period still selling is projected at the last 28 days’ pace. Enter what arrived to check the estimates against it.
All amounts are estimates from RevenueCat: each sale’s price less the store’s commission and tax as RevenueCat reports them, converted to USD on the day of the sale. App Store Connect and Play Console can differ by a few percent, for local taxes, their exchange rates and refunds; see What the estimates leave out, below.
Cash by month
Ad spend going out, Meta’s, Google Ads’ and Apple Ads’ where connected, against store payouts coming in, in the month the money moves. Ahead, the campaigns running now, each at the share of its daily budget it spent lately, paused ones left out, and the stores keep their last 28 days’ pace.
Calendar
The days money moves: each store’s paydays, the days its periods close, and ad spend day by day.
Other money in and out
Optional. Loan and credit-line payments, money you expect in, and fixed costs such as API bills, on the days they fall. They count in Cash by month and show on the calendar.
Add an item
How the stores pay
The rules behind the dates above.
App Store
- Paid per fiscal month of 4 or 5 weeks, Sunday to Saturday, on the day Apple’s fiscal calendar sets: usually 33 days after the month ends, and by its terms within 45.
- Paid for each country or region only above its minimum payment threshold; less carries over to the next payment.
- Paid in your bank account’s currency, at Apple’s exchange rate.
This fiscal year’s calendar
Google Play
- Paid per calendar month: orders, refunds and chargebacks from the 1st to the last day are paid out around the 15th of the next month.
- No deposits on weekends or bank holidays, so a 15th on a weekend moves to the Monday; banks then take a few days to credit it.
- Paid once the balance reaches the minimum: US$1 for local-currency payouts, US$100 for USD wire transfers.
What the estimates leave out
From RevenueCat’s prices. Each sale’s USD price less the commission and tax RevenueCat reports for it. The stores convert at their own rates and pay in your bank’s currency, so a payout lands a little off the estimate.
Since Bridge started recording. Sales before Bridge received them are missing, so a period marked partial is short. Refunds are counted from 26 October 2026, when Bridge began recording them.
Tax withheld. Whether Apple or Google withhold tax from your payouts depends on where you are based and the tax forms you filed in App Store Connect and Play Console. If they do, set the share under Take off estimates.
Months close in Pacific time, the stores’ home zone, so a sale in a month’s last hours may count in the next one. Apple publishes its calendar a year at a time: Bridge has it through September 2027, and later months follow its usual pattern until it is added.
Match rate
Of the installs that had an ad click to match, how many Bridge connected to one.
How match rate is calculated, and what ambiguity means
Match rate is measured against installs that had a click to match, not against every install. Organic installs have no ad click and can never match, and Google Play's pre-launch report installs every uploaded build from Google's own datacenter — counting either as a miss would blame the matcher for having nothing to attribute. Those are shown asNo click available. The whole-funnel number to
watch is match % in the table below, which is installs over
clicks.
A third-party attribution SDK running alongside Bridge does not affect any of this: it never touches matching, delivery, or these counts.
Ambiguous installs are counted separately from unmatched on purpose: two clicks scored within the confidence margin of each other, so attributing either would be a coin flip. A rising ambiguity share is the earliest signal that the scorer is losing discriminative power — usually after an iOS release — and the cue to re-tune weights with
tools/replay.ts.
Meta delivery
Conversions sent to Meta through the Conversions API in this period.
Attribution resolution
How each conversion was tied back to its source.
Reading the unattributed share
How each conversion resolved back to a click. A largeunattributed share is the honest read on revenue the bridge
failed to connect — compare against SKAdNetwork as an independent
baseline. If this reports far more installs than SKAN sees, suspect
false positives rather than celebrating.
Relay and VPN traffic
Ad clicks that arrived through iCloud Private Relay or a VPN.
Why relay traffic mostly goes unmatched
Traffic behind iCloud Private Relay or a VPN has no usable IP signal, so it is matched on a 2-hour window at a much higher threshold and mostly goes unmatched by design. A high relay share is the usual explanation for an otherwise puzzling match rate.Attribution
Resolved by Apple from the AdServices token, through the Bridge SDK and RevenueCat.
Where these numbers come from
Apple resolves an install to a campaign, ad group and keyword from its AdServices token. The Bridge iOS SDK 0.7 or later sends that token with the install, and Bridge asks Apple; RevenueCat’s Apple Search Ads integration does the same and writes the result onto the subscriber, which reaches Bridge with every webhook. A day after Apple answers, Bridge fills in a RevenueCat customer RevenueCat left with no source.The media source is checked before anything is stored. Those reserved fields are shared by every attribution source RevenueCat knows about, so reading them without that check would file Meta campaign names as Apple keywords.
Keywords by revenue
Subscribers, trials and revenue per keyword in this period.
Why this table is the point of building any of it
Apple reports spend and installs and knows nothing about what happens afterwards. AppTweak and Sensor Tower model search volume and know nothing about your subscribers. The join between an Apple keyword and the revenue it actually produced only exists where both halves already live, which is here — and it is what turns bidding on cost per install into bidding on the business.Spend, cost per trial and the return of each keyword are on Automation → Apple Ads, which sets Apple’s spend against these users.
Checklist
Every tick is evidence, not configuration: a webhook counts once an event has arrived, the SDK once a device has claimed. Hover a cell for details.
Apps
Registered apps, one row per platform.
Ad links
The link for each ad's Website URL field, per platform, for the app picked at the top. Add a custom product page or custom store listing to get a link that opens it.
Custom pages, macros, and why the host matters
Custom pages. A custom product page (App Store) or custom store listing (Google Play) gets its own ad link. The page id rides in the link asppid= or listing= and Bridge passes it on to
the store, so attribution works exactly as for the default page. Because
the id is in the link itself, removing a page here never breaks an ad
already using it. Find the id in App Store Connect under the app's Custom
Product Pages, or in Play Console under Custom store listings; pasting the
whole page link works too.
The
{{...}} tokens are expanded by Meta at click time —
that is where ad-level granularity comes from, independent of Meta's
own reporting. An ad configured without them delivers the macro
literally; those are discarded rather than stored, so a
misconfiguration shows up as missing data rather than as a campaign
called {{campaign.name}}.
Only
bridgemmp.com is offered. bridge-mmp.com
still resolves so an old link never breaks, but ad URLs must use one
canonical host: event_source_url is stored verbatim from
whichever host the click arrived on, and splitting live traffic across
two hostnames splits Meta's view of the domain with it.
Campaign must be Sales objective with a website destination. App Promotion campaigns on iOS are restricted to SKAdNetwork and cannot carry the click id at all — that limitation is the reason this system exists.
RevenueCat webhook
Paste into the RevenueCat project's webhook settings, with the Authorization header below. For the app picked at the top.
In RevenueCat, add the header Authorization with the app’s value from its row. The value is copied, never shown.
One webhook per RevenueCat project, not per app
RevenueCat configures webhooks per project, and a project usually holds both the iOS and Android app. Point it at either slug: when the event's store contradicts that app's platform, Bridge hands it to the configured sibling, so a Play purchase cannot be reported to the iOS dataset. Add an app links the two platforms for you, and so does + Android or + iOS when the second one comes later.The
Authorization header value is the shared
RC_WEBHOOK_AUTH secret — the same string for every
app. It is not shown here; Bridge never displays a secret it holds.
A debug app has its own slug, so a sandbox project can point at
-dev and never touch production attribution.
RevenueCat’s own fields
Bridge always writes itsbt_ attributes on a subscriber an ad brought: network,
campaign, ad set, ad, keyword and match type. With Also write RevenueCat’s own fields
on, it writes RevenueCat’s reserved ones as well: $mediaSource (Meta Ads,
Google Search or Google UAC), $campaign, $adGroup,
$ad and $keyword, which RevenueCat’s own charts and integrations
split revenue by. Leave it off when another tool, such as an MMP or one of RevenueCat’s
integrations, writes them already: two writers would overwrite each other. It needs the
Bridge SDK 0.6.1 on iOS or 0.3.1 on Android, or later, and applies to subscribers from then on.
RevenueCat subscriptions
Renewals ahead for the Claude connector, annual plans bought before the webhook included: who renews when, and who has turned auto-renew off. With Charts & metrics, RevenueCat’s own charts too. A key for the app’s RevenueCat project, kept encrypted: it reads, writes customers’ attributes, and, for Plans, price tests and products made through the connector, products and offerings.
Loading…
Making the key in RevenueCat
In RevenueCat, open the app’s project, then Project settings → API keys → + New secret API key, choose API version 2, and open each group with the arrow beside it:- Customer information: Customers Configuration on Read & write; Subscriptions, Purchases and Invoices Configuration on Read only. Bridge writes onto customers: RevenueCat’s own Media Source, Campaign, Ad Group and Ad where still empty, a Google Search install’s keyword, and a price test’s offering. Set each line, not the group’s own setting, which switches all four: a key that can write purchases or subscriptions is more than Bridge needs if it ever leaked.
- Project configuration: Products, Packages, Offerings and Entitlements on Read & write, the rest on Read only: for Plans, price tests, and new products made through the Bridge connector (Guides → Plans, step A).
- Charts & metrics: Read only, for RevenueCat’s charts in the connector.
- Everything else: No access.
Meta
What is sent to Meta, and where. For the app picked at the top.
Meta connection
Fill these in once the dataset exists. Delivery stays off until a dataset is set and verified, so saving here cannot start sending on its own.
The secret
name, never the token itself — the token goes in with
wrangler secret put <name>, or is pasted in Guides, Meta.The access
token itself is pasted in Guides, Meta, under Your values: it is stored encrypted and never shown again.
Why collapsing events concentrates the signal
Meta optimises against a single event. Sending trial starts asStartTrial, paid conversions as Purchase and
Subscribe splits one commercial moment three ways, and
each event needs its own ~50 weekly conversions to leave the learning
phase. Collapsing them into Purchase concentrates the
signal — and it is the usual default for a subscription app.
The trade-off is that you lose the trial/paid distinction inside Meta. The full breakdown is still in this console and in RevenueCat, so it costs reporting detail there, not here.
Meta Ads API
Your ad accounts, read by Bridge: spend for ROAS, and campaigns created from Claude — always paused, set up for the web funnel.
Checking…
Where new campaigns go
For the app picked here: the ad account, Facebook Page and Instagram account that campaigns made from Claude use.
App stores
App Store Connect and Google Play, read by Bridge for your plans: each subscription’s price, state and who holds it, on Plans; and, for a price test, the products it makes. How: Guides → Plans.
App Store Connect
Checking…
Google Play
Checking…
Google Ads API
Your Google Ads account, read by Bridge as a service account: each App or Search campaign’s spend and results, filed under the app it promotes; new campaigns from Automation’s planner or Claude (always paused), their budgets, keywords, ad text and on/off. Keyword search volumes need Google’s Basic access: Guides → Google Ads, step F.
Checking…
Basic access: brand verification first
Explorer covers everything Bridge does, and keyword figures come through Bridge’s own project until then. Basic raises the limit from 2,880 to 15,000 operations a day and has Keyword Planner answer from your own project. Google grants it only to a project whose brand is verified: until then Upgrade access level on the API’s Overview page says Brand Verification needs to be completed on your account to be approved for Basic access, and an application is refused.
As in Google’s brand verification guide, all in the Google Cloud project Bridge is connected with:
- In Cloud Console, open Google Auth Platform (APIs & Services → OAuth consent screen). If it is not set up yet, Get started: an app name, a support email, External audience and a contact email, then Create. Bridge does not sign anyone in through it; Google asks for it for Basic all the same.
- Audience: the user type must be External (Make external if it says Internal) and the publishing status In production (Publish app, then Confirm). Other Google pages say an Internal or Testing app needs no verification; for Google Ads API Basic access it does.
- Branding: your app or company name, a logo that is your own brand’s (120×120 pixels), your home page, privacy policy and terms of service on a domain you own, and that domain under Authorized domains. Save.
- Verify the domain as yours in Google Search Console, with the Google account that owns the Cloud project.
- Back on Branding, under Verification status at the top right, Verify branding. Often it takes minutes; when Google’s team looks at it, a few days. If it lists issues, fix them, wait the day it asks for, then View issues → I have fixed the issues. If it flags your logo wrongly, choose I believe the issues found are incorrect and say where the logo is used.
- Once verified, Publish branding. Branding then shows the project as brand verified.
- On the Google Ads API’s Overview page, Upgrade access level: the next level says Basic, with no brand warning. Apply for access: a short form on what the tool does. Google reviews it, usually within a few business days, and emails the outcome.
Once granted, set Access level above to Basic. In the guide: Guides → Google Ads, step F.
Apple Ads API
Your Apple Ads account, read by Bridge as an API user of yours: each campaign’s, ad group’s and keyword’s spend, taps and installs, and the searches that showed your ads, filed under the app they promote; bids, keywords, negatives, budgets and new campaigns from Automation and Claude. Which keyword brought each user, Apple tells Bridge for installs of app versions with the iOS SDK 0.7 or later.
Checking…
GA4 → RevenueCat · App campaigns
Google’s own answer to which App campaign and ad group brought each user, read from Google Analytics and written onto the RevenueCat customer. iPhone installs get it only here; on Android it fills the clicks whose Play referrer names no campaign, and adds the ad group. For the app picked at the top.
Writes $mediaSource,
$campaign and $adGroup onto each RevenueCat
subscriber that Google Ads acquired, read from GA4. It needs: Firebase Analytics in the
build people have (not only other Firebase products), the app passing its Firebase id to
RevenueCat (one line, below), Google Analytics’ daily BigQuery export with
first_open not excluded and every live stream ticked, and BigQuery Data Viewer
and Job User for the Google Ads API’s service account. The dataset appears the day after
the first events. It starts in dry run — it reports what it would
write, and writes nothing, until you switch it to write.
Step-by-step: Guides → Google Ads → GA4.
Where to find them: the project id is the one Firebase shows in Project settings → General
(the export goes to the Firebase project). The dataset is analytics_ followed by the GA4 property id, from
Google Analytics → Admin → Property details; it is listed under the project in
BigQuery a day after the link is made.
Worker secrets instead (owner only)
Left empty, the Google Ads API’s service account reads BigQuery and the RevenueCat key on Setup writes the customers.
Data retention
The nightly clean-up of old attribution data, delivery records and history.
What is kept, and for how long
Scrub signals early, keep results. Device signals exist only to match a click to an install, inside a window of at most 24 hours; after that they are useless and are removed. The outcome — which campaign won — is what every report depends on, so it is kept.Clicks nothing points at: deleted at 90 days. Clicks an install or a conversion points at: fingerprint removed at 90 days, campaign kept. Raw IP addresses: 30 days, long enough to reach Meta with late trial conversions. Install device signals: 95 days. Installs, purchase and renewal events, subscription state, Google Ads links: 760 days, so two full years can be compared.
The pings sent to Meta for each ad click and matched install: 90 days, with their clicks (Meta already has them). Meta’s reply to each delivery: 30 days. AI analyses: 760 days, with the numbers each was given trimmed to campaign names after 90 and failed runs removed after 30. Budget changes Bridge applied, the audit log and Meta spend: 760 days; refused budget changes: 90. Google Ads click lookups, daily campaign figures, price-test groups and Search Ads attribution: 760 days; the record of each conversion sent to Google: 120 days.
Results are kept about as long as paid MMPs keep raw data (two years); device signals and IPs go much sooner than theirs.
Danger zone
Permanently deletes every click, install, event and subscription-state row. App registrations and the ASN registry are left intact, so live traffic keeps working. There is no undo and no backup.
AI plan · Google Ads
What to do next with this app’s Google campaigns. Nothing changes until you apply a step, or the autopilot does within your goals below. Each plan reads the last two weeks against the last eight, so a few unusual days do not decide it.
Goals and autopilot · Google Ads
What the Google plan and its autopilot follow, apart from Meta’s. Amounts in the Google Ads account’s currency.
Plan a new campaign · Google Search
The AI drafts a Search campaign for this app in the language you pick: its keywords, the searches it left out and why, the ad, the budget and the countries, from the app’s store listing, the searches its campaigns have shown for and your goals. Google checks it before anything is made. Change what you like, check it again, then create it paused: nothing spends until it is switched on.
Campaigns
Google’s figures for the period picked, beside what Bridge saw the users each campaign brought pay.
How these are counted
A conversion is what the campaign optimises for: an install, or an in-app action. Google’s value is its own worth of those conversions, before the store’s share; ≈ proceeds is that value in dollars after the share your stores keep, an estimate. Bridge’s column is its own count: the Android users whose Play install referrer names the campaign (or carries Google’s click id while it is the app’s only Google campaign spending), a Search campaign’s iPhone users matched through Bridge’s link, and everything they have paid since, counted from the day they installed. An iOS App campaign gives Bridge nothing to match.Keywords · Search campaigns
Each keyword over the last 30 days: Google’s spend, clicks and conversions, beside what Bridge saw of the people its clicks brought. Below, the searches that showed your ads: what the AI adds as keywords, or blocks.
Searches that showed your ads
Ad text
Each ad’s headlines and descriptions, Google’s rating of each, and what each line brought. Open a campaign to see or edit its lines.
How lines are judged and changed
Google rates how a line draws attention (Best, Good, Low, or Learning while it gathers data), not what the people it brings pay. The AI and the autopilot replace a Low line only when its value for its spend is below most of the ad’s too, and keep one that earns its place; they add lines to an ad with room for more. Three changes an ad in two weeks at most. Edit rewrites any line yourself. Each change keeps the line’s pin, can be undone below, and is reviewed by Google before it shows; a new line is rated after a week or two.Changes Bridge made in Google Ads
Every budget, pause, switch-on, keyword and ad text change made from here, by the autopilot or from the connector, and every step refused. A keyword or text change can be undone. The AI sees them too, so it knows what was done and when.
AI plan · Apple Ads
What to do next with this app’s Apple Ads: bids, keywords, blocked searches and budgets, judged on what the users each keyword brought go on to do. Nothing changes until you apply a step, or the autopilot does within your goals below. Each plan reads the last two weeks against the last eight, so a few unusual days do not decide it.
Goals and autopilot · Apple Ads
What the Apple plan and its autopilot follow. Amounts in the Apple Ads account’s currency.
Campaigns
Apple’s figures for the period picked, beside what Bridge saw the users Apple credits each campaign with do: trials, payers and proceeds. The keywords below follow the same period.
Keywords
Each keyword over the period picked above, most spent first: its bid and Apple’s suggestion, Apple’s taps and installs, and Bridge’s trials and proceeds. Below, the searches that showed your ads over 30 days.
Searches that showed your ads
How these are counted
Apple’s spend, taps and installs are its own, in the account’s currency and time zone, read hourly. Bridge’s figures come from Apple’s attribution answer for each install of an app version with the Bridge iOS SDK 0.7 or later (and RevenueCat’s Apple Search Ads fields, where the app turns them on): the users it credits a keyword with, and every trial and payment of theirs since. Searches Apple counts too few of to name are left out of its list.Plan new campaigns · Apple Ads
The AI drafts the classic set for this app from its App Store listing and Apple’s keyword suggestions: brand, category, competitor (only if your notes allow it) and discovery, each with its ad groups, keywords, bids and blocked searches. Bridge names each one App - Kind - countries - App Store id. Look it over, then make them: they are made paused, and nothing spends until you switch one on.
Changes Bridge made in Apple Ads
Every bid, keyword, blocked search, budget and on/off change made from here, by the autopilot or from the connector, every campaign made, and every step refused. Each can be undone.
AI plan
What to do now, from the numbers below and your goals. Budget steps can be applied to Meta from here: Bridge checks each one against your limits and Meta’s live budgets first. The plan weighs the period picked against the last eight weeks, so a few unusual days do not decide it.
Goals and autopilot
What the AI plan and the autopilot follow
By platform
Users who clicked in this window and what they have paid so far, against what was spent in it: USD after the store’s cut and each country’s VAT. Android is measured exactly through the Play install referrer; iOS matching loses some paid users, so iOS is shown as a range.
Campaigns
Projected adds trials still running, at the app’s own conversion rate. Paused campaigns are judged on their history. Hover a verdict for the rule behind it.
Trends
Each campaign week by week, by when its users clicked, up to yesterday: where it is heading, which an average over any window hides. Direction compares the cost per trial or purchase over its last three weeks of spending with the weeks before.
Ad fatigue
Whether each running campaign’s main ads are wearing out, from their last eight weeks at Meta. An ad needs five weeks of delivery before it can be judged, and a recent budget raise or new ads hold the call. Slipping is 10 to 20% under its usual click-through, Wearing out is 20% or more. The AI asks for new creatives against fatigue only at Wearing out.
Occasions
Seasonal ads judged as one group: the users who clicked them between the occasion’s start and its end, against what those ads spent, beside the other ads in the same ad sets over the same days. Ads are added to one through the Bridge connector (add_meta_ads with an occasion), and Bridge pauses them when it ends.
Add an occasion
Countries
Meta’s spend in each country against what users who clicked there paid, and the share of the price kept after store fees and tax.
Changes Bridge made
Every budget, pause and restart applied from here or by the autopilot, seasonal ads switched on and off, and every step refused. The AI sees this list too, so it knows what was done and when. Kept for 13 months, refused steps for 30 days.
How these numbers are worked out
By cohort. A campaign is credited with the users who clicked its ads inside the window, and with everything they have paid since: first payments, trial conversions and renewals. Users who came before the window, or not from an ad, are not counted, and neither are renewals still to come. Users from the last days have had little time to pay, so short windows read low.
Proceeds are RevenueCat’s USD price times the take-home share it reports, which already removes the store commission and the country’s tax. A purchase in France keeps about 72%, one in the US about 85%.
Projected adds each trial still running, valued at what this app’s trials convert at. A trial is running until its own end date from RevenueCat, or, for trials recorded before Bridge kept that, until the delay by which nine in ten of the app’s trials have converted.
iOS estimate. Android tells Bridge exactly which installs came from a link, so the share of new-subscriber money that is truly organic is known there. Assuming iOS has the same share, whatever iOS earned beyond that, and beyond what was matched, was paid but missed, and it is spread over the iOS campaigns by their clicks. It is an assumption, so iOS is shown from matched to estimated.
Verdicts follow your goals: keep from your target ROAS, scale above it, watch below, cut well below. A campaign in its first 7 days, still in Meta’s learning phase, under $75 spent or with fewer than 8 trials and purchases is too early. A paused campaign is judged on everything Bridge has on it, up to 90 days: restart, retest small, or leave paused.
Connect
Totals only, never a user or a click. Claude reads by default; creating, editing and deleting paused campaigns is allowed per connection below. Nothing to copy: Claude signs in to Bridge, and you allow it.
- Add to Claude. Opens claude.ai with Bridge filled in. Click Add, then Connect.
- Allow. A Bridge page asks whether Claude may read your data. Signed in here, it is one click.
- Ask. In a chat with Bridge switched on, for example “How did my campaigns do this week?”
Your connections
Each is yours, not this workspace’s: it works in every workspace you are in, as you can there. Claude works in the workspace you last opened here, or the one an app or your question names. Disconnecting one ends its access at once; it can connect again the same way.
Claude Code
For the claude CLI. It signs in the same way the first time it uses Bridge.
Or use a key instead of signing in
For a client that can only send a fixed header. The key goes straight to the clipboard and is never shown; paste it as the value of anAuthorization header.
wrangler secret put MCP_TOKEN.
What Claude can see and do
Campaigns, ads, install sources, match rate, Meta delivery, Google Ads, Apple Search Ads keywords, ad links and the AI budget plan, as totals. Never an individual install, click, IP or RevenueCat user. With Create and edit ticked it can upload ad media, create Meta and Google Ads campaigns (always paused) and edit them, change a daily budget, and switch a campaign or ad on or off: each is shown to you first and made only once you say yes in the chat. With Delete, it can remove ones that are not running. Every call is recorded in the audit log, which is where last used comes from.Figma, for ad creatives
With a Figma token, Claude can turn a link to a Figma frame into an ad image and upload it to Meta, in one chat. Read access to file content is all it needs.
In Figma: Settings, Security, Personal access tokens, with File content read access. It is checked with Figma, stored encrypted, and never shown again.
Billing
Your plan, your card and your invoices, in US dollars. Each month’s use is charged at the start of the next.
Workspace
Its name, as everyone in it and its invitations see it.
Invite someone
Bridge is invitation only. An admin sees and changes everything in this workspace, its settings and people included. Anyone else sees only the apps you choose for them, as a viewer (numbers and the AI plan) or a manager (who can also apply budget steps, edit ads and occasions). Someone who already has a Bridge account is invited the same way, and joins by signing in.
People
Everyone in this workspace. Disabling someone signs them out at once; removing takes them out of this workspace, with their Claude connections here.
Open invitations
Each link works once and expires after 7 days. A new invitation to the same address replaces the old link.
Needs attention
Money that is late, access that is paused, and what is about to change.
Clients
Each client’s plan, prices and limits. Edit one to give it free access for a time, a plan of its own, a discount, or more room. Diamobyte is yours, so it is not here.
| Loading… |
Invoices
The latest 200, newest first. Refunds and disputes are handled in Stripe: each invoice links there.
Month by month
By the month of use each invoice charges for.
Standard prices
What clients pay unless they have a plan of their own. See, as you type, what each workspace would pay for its real months. Save to keep them: the next month’s bills use what is saved then.
Country tiers
Per attributed install, by the country it came from: two-letter codes, such as US, GB, DE. A country in two tiers takes the first one’s price.
Try your own numbers · for a client you are quoting
When an invoice goes unpaid
For every client, unless one has its own days to pay.
Paused: its people cannot use the workspace or its Claude connections, except its admins, who can still pay. Analyses, spend sync and scheduled ad starts pause. Tracking links, the SDK and postbacks keep running, so live ads keep working and no data is lost, and ads scheduled to stop still stop. Access comes back as soon as the invoice is paid.
History
The latest changes to plans, prices and invoices, and who made them.
Workspaces
Each client’s own Bridge: its apps, people, Meta connection, goals and payouts. You can open any of them and work in it as an admin.
| Loading… |
Requests for a higher app limit
From people at their limit. Approving sets their new limit; they are emailed either way.
Usage
Each workspace’s production apps over the last 30 days, and its Claude analyses this month.
People
Every account, in every workspace. Disabling someone signs them out everywhere; deleting removes the account from all its workspaces.
Settings
For the whole service.
Profile
Signing in
With Google, or with your email and a password. Either works once set up; keep at least one.
Delete your account
Your account, your access to every app and your Claude connections go at once, and cannot be brought back. An admin can invite you again.
As the owner of Bridge, your account cannot be deleted.
Meta — end to end
Your values
Bridge carries a Meta ad click through the App Store or Google Play and back into the app, so conversions reported to the Conversions API arrive with the original click attached and Meta can populate Campaign, Ad set and Ad on them. This is the whole sequence, in the order it has to happen.
Before you start
- A Meta dataset for this app, in Events Manager, created as a Web data source. One dataset per app per platform — iOS and Android should not share one, or you cannot tell the two apart in reporting.
- A Conversions API access token generated from that dataset.
- RevenueCat configured for the app.
The sequence
Events Manager → Connect data sources → Web, not
App → Next. Name it so the platform is obvious, e.g.
MyApp iOS. If it asks for a website, give
bridgemmp.com; if it asks how to connect, choose the
Conversions API and skip installing pixel code — Bridge
sends the events. Copy the numeric dataset id.
Then Settings → Conversions API → Generate access token, from this dataset. Keep the tab open; the token is shown once.
Setup → Apps → Add an app. It registers every platform the app ships on, production and debug, and shows each one's client key. The dataset id goes in afterwards, under Your values above (or Setup → Meta). Delivery to Meta starts off.
events_received: 1 and no
error. Check the id against Events Manager yourself.Paste it into Access token under Your values above. It is stored encrypted and never shown again. Or, from a terminal, as a Worker secret, which takes precedence over a pasted one:
wrangler secret put META_TOKEN_MYAPP_IOS --config apps/bridge/wrangler.toml
Then press Verify. It catches a missing token and a token Meta rejects — which is what it can prove, and no more.
RevenueCat → Integrations → Webhooks. URL:
https://bridgemmp.com/hooks/revenuecat/<app-slug>
Set the Authorization header to the app’s value: copy it from the app’s row in Setup → RevenueCat webhook.
sibling_slug) so Play Store
events are routed to the Android app; without that link Play purchases are reported
to the iOS dataset, credited to an install that does not exist,
and the numbers still look plausible.Sales objective, website destination. Copy the URL from
Setup → Ad links and paste it into the ad's Website
URL field, macros included. The {{...}} tokens are
expanded by Meta at click time and are where ad-level granularity
comes from, independent of Meta's own reporting.
- Keep Skip the install page unticked while you create or change ads, until Meta has approved them.
- Once they are approved, tick it if you want taps to go straight to Google Play, one step fewer for people.
- Untick it again before you add or change ads, and tick it once they are approved.
iOS: the BridgeAttribution Swift package.
Android: com.bridge:attribution. Configure with the
slug and client key from step 2, and pass the RevenueCat app user
id as soon as you have one — the webhook resolves purchases
through it.
On a debug build, open your ad URL, install through the store,
and check the console. You want an install with a
match_method. Then make a sandbox purchase and check
Events Manager → Test Events.
probabilistic or nothing, the referrer is not
surviving — stop and look before running budget.The moment the dataset and the token are both saved, Bridge asks Meta whether it accepts the token, and if it does, switches delivery on: new conversions go to Meta from then on. If Meta refuses the token, delivery stays off and the card says why.
To pause, untick Send attribution to Meta in Setup → Meta. A paused app stays paused when its dataset or token is changed; Turn on delivery in the card resumes it. Use the console, not SQL — see the first troubleshooting entry.
Setup → Meta Ads API: paste a system user token and pick each app's ad account, Facebook Page and Instagram account. Bridge then reads spend from Meta every hour, so ROAS in Campaigns is real, and Claude can create campaigns for you. Getting the token is its own sequence: Connect the Meta Ads API, below. For higher rate limits, Move to Full access.
A campaign created from Claude is always this funnel and always paused: a Sales campaign whose ads open the app's Bridge link with its tracking parameters, optimised for Purchase on the app's dataset, shown only on the platform the link leads to. It is refused for an app whose delivery to Meta is off. You switch it on in Ads Manager, or tell Claude to, once it has shown you what would go on and the daily spend.
Connect the Meta Ads API
A different token from the dataset’s in step 1. That one only sends conversions to one dataset. This one belongs to a system user of your Meta business, and it is what lets Bridge work in your ad accounts: read each campaign’s spend, so ROAS in Campaigns is real; create campaigns and upload images from Claude; and change budgets from Automation. One per workspace, pasted once in Setup → Meta Ads API.
The Meta app is your own. Each business using Bridge creates one, connected to its own business portfolio, and it reaches only the ad accounts you assign to it. Meta does not charge for it.
What you need first
- A Facebook account with full control of your business portfolio. No portfolio yet? Step A can create one.
- An ad account in that portfolio, with a payment method.
- A Facebook Page for each app. To run ads as Instagram too, its professional account connected to that Page.
Open My Apps and click Create app. First time here? Register as a developer first. If Meta says Page Not Found, you are signed out: sign in and open the link again.
Meta then asks, in order:
- App details: a name, such as Bridge, and a contact email.
- Use cases: tick Create & manage ads with Marketing API. An older form asks for a type instead: pick Business, then add the Marketing API product.
- Business: pick your business portfolio, or Create a business portfolio. Do not skip this: a system user only works with an app in its own portfolio.
- Requirements and Overview: Next, then Go to dashboard.
Then, in the sidebar, open Use cases, open the ads use
case, and add the optional permission pages_manage_ads,
plus business_creative_management if it is listed. A
permission left out here may not be offered in step E.
Leave the permissions on Standard access. That is enough, and you do not need Advanced access, because Bridge only works in ad accounts your own business owns.
In the sidebar, App settings → Basic:
| Privacy policy URL | https://bridgemmp.com/privacy |
| Terms of service URL | https://bridgemmp.com/terms |
| User data deletion | If it asks, the same privacy page: it says how to ask for deletion. |
| App icon | 1024 × 1024, without Meta’s logos or the words Facebook or FB. |
| Category | Business |
| App purpose | Yourself or your own business |
Save, then click Publish in the sidebar and publish the app. The sidebar then shows Published. Some dashboards call this switching from Development to Live.
Open Business settings → Users → System users and add one. Name it Bridge, role Admin. A system user belongs to the business, not to a person, so its token keeps working when someone leaves or changes their password.
Already have an admin system user? Use that one: a new app allows only one, and one other. Meta’s help: Add system users.
Open the system user → Assign assets, and give it:
- the Meta app from step A;
- your ad accounts, with Manage campaigns (or Full control);
- each app’s Facebook Page, with Create ads;
- each app’s dataset.
Ads run as the Page and its Instagram account, so connect each Instagram account to its Page: Bridge finds Instagram accounts through their Pages.
On the system user, Generate new token. Pick the app from step A, expiry Never, and these permissions:
ads_management ads_read business_management pages_show_list pages_manage_ads
Meta may tick pages_read_engagement as well: leave
it on. Add business_creative_management too if Meta
offers it: it lets uploaded images go to your Media Library, in a
folder per app. Copy the token; Meta shows it once.
To check a token before pasting it, Meta’s Access Token Debugger shows its app, its permissions and when it expires.
Setup → Meta Ads API → paste → Connect. Bridge checks it with Meta, lists the ad accounts it reaches, stores it encrypted, and never shows it again. Then, for each app, pick the ad account, Facebook Page and Instagram account its campaigns are created in, and Save. Spend arrives within the hour.
- Setup → Meta Ads API lists your ad accounts. Once Meta has answered, it also shows your access tier; a new app shows Limited.
- Campaigns shows spend within the hour, if your ads have spent.
Everything now works on Limited, within its low rate limits. The next section lifts them.
When Meta refuses. Bridge passes Meta’s refusals on in plain words; this is what each one means.
| What you see | What to do |
|---|---|
| Page Not Found on My Apps | You are signed out of Meta. Sign in and open the link again. |
| No more system users can be added | A new app allows one admin system user and one other. Use the one you have (step C). |
| The app or a permission is missing when you generate the token | The app is not in the system user’s portfolio or not assigned to it (steps A and D), or the permission was not added to the use case (step A). |
| This token reaches no ad account | The ad account is not assigned to the system user. Assign it with Manage campaigns (step D), generate a new token, and paste that. |
| Every ad account this token reaches is connected in another Bridge workspace | An ad account belongs to one workspace at a time. Disconnect it in the other workspace first. |
| Your Meta app is in Development mode | Publish it (step B) and try again. |
| The token lacks a permission | Generate a new token with every permission in step E, with the ad account, Page and dataset assigned, and paste it. |
| The Meta Ads API token is no longer valid | It expired, or someone revoked it. Generate one with expiry Never and paste it. |
| User request limit reached, or too many calls from this ad account | Meta’s rate limit. Calls to that ad account stop for 5 minutes on Limited and 1 minute on Full, then work again. If it keeps happening, move to Full access. |
| A Page or Instagram account is missing from the lists | The Page is not assigned to the system user, or the Instagram account is not linked to the Page (step D). Fix it in Meta and pick the ad account again. |
| This ad account cannot use the chosen Instagram account | Pick another in Setup → Meta Ads API, or none: the ad then runs as the Page alone. |
| Ads reaching the EU need a beneficiary and payer | The EU’s Digital Services Act. Set who benefits from and who pays for the ads in the ad account’s settings in Ads Manager, once. |
| Images go to the ad account, not the Media Library | Meta refused Media Library uploads with this token. Generate
one with business_creative_management if Meta
offers it. Bridge tries the library again each day. |
To stop: Disconnect in Setup → Meta Ads API deletes the token from Bridge. Campaigns already created stay in Meta, as they are. Revoke the token in Business settings too if it is not used anywhere else.
Move to Full access
A new app starts on Limited access, which Meta says is for development and not for live advertisers. Full takes one App Review, and is worth having before you create or start many ads at once. These limits apply to each ad account:
| Limited | Full | |
|---|---|---|
| Changes before Meta stops your calls | About 20 in a few minutes | About 3,000 |
| How long calls stop for | 5 minutes | 1 minute |
| Ad management calls an hour | 300 + 40 per active ad | 100,000 + 40 per active ad |
| Reporting calls an hour | 600 + 400 per active ad | 190,000 + 400 per active ad |
Full access changes only these limits. Your permissions stay on Standard access, as in step A.
When you can ask: once the app has made at least 500 Marketing API calls in the last 15 days, with under 15% errors in the last 500. Bridge’s hourly spend sync counts towards this, and Meta’s form shows when you have reached it.
In your app’s dashboard, open Permissions and features, find Marketing API Access Tier and click Upgrade. Older dashboards list it under App Review. This only adds it to your submission: nothing goes to Meta until step 6.
Paste this into Tell us why you’re requesting, tick the agreement, and Save. Put your company’s name in, and delete anything you do not use, such as the automation sentence.
[Your company] runs advertising for its own mobile apps and uses Bridge, an attribution and advertising console, to measure and manage it. This app connects Bridge to ad accounts in our own business portfolio, and it reaches no other business's accounts. How we use the feature: through a system user, the app reads ad performance reports (spend, impressions, clicks and results by campaign, ad set and ad) and manages our ads: it creates campaigns, ad sets and ads, uploads creatives, changes budgets, and starts or stops ads on planned dates. AI-assisted automation can also adjust budgets within fixed rules. Every change is logged, and budget changes can be undone. The data is used only to measure and manage our own advertising. We do not sell it or share it with data brokers. How it adds value: our team sees ad spend beside installs, trials and revenue in one place, so we can move budget to what works, spot ads wearing out, and launch seasonal ads on time without manual work in each account. Why it is necessary: Limited access is intended for development, and normal production use exceeds it. Planned launches and scheduled starts and stops of many ads must happen on time, so we need Full access for our live advertising.
These are your own statements to Meta, so check each one is true for your business. Meta explains them here.
Do you have data processors or service providers? Yes. Leave Anthropic out if you do not use Automation, and add any other company of yours that can see this data.
Diamobyte Software FZCO (operates Bridge: attribution and ad management); Cloudflare, Inc. (hosting, database, storage and email delivery); Anthropic, PBC (AI analysis of advertising figures by campaign and ad)
| Question | Answer |
|---|---|
| Category of services | Diamobyte: Analytics and measurements, Advertising and IT solutions and services. Cloudflare: IT solutions and services. Anthropic: Analytics and measurements. |
| Who is responsible for the data | Your company’s exact registered name, as in your Meta business details. Not Bridge: you decide how your advertising data is used. |
| Country | Where your company is registered. |
| National security requests in the past 12 months | No, unless you have given authorities this kind of data. |
| Policies for requests from public authorities | Tick the ones you really have in writing. If you have none, None of the above. |
App settings → Basic → Add platform → Website. Site URL:
https://bridgemmp.com/
Meta’s Reviewer instructions stay locked until this is saved.
Meta’s reviewers need a way into Bridge. In Users → Invite someone, invite an email address you control, with Chosen apps, as a viewer of one app. Open the invitation in a private window and set a password you use nowhere else.
A viewer can change nothing, but it sees that app’s real figures, so choose the app accordingly. Never give Meta your own login. Meta asks for test access to stay active for a year.
On the Website platform, Add or update instructions:
Where can we find the app?
https://bridgemmp.com/console
Bridge is a private console, and access is by invitation only. Use the test credentials provided in this form. The test account is read-only and shows our real campaign figures, so please keep them confidential. 1. Go to https://bridgemmp.com/console 2. Sign in with the test email and password. Do not use "Continue with Google". 3. Open Campaigns in the sidebar. It shows spend by campaign, read through the Marketing API, beside installs, trials and revenue. 4. Open Automation in the sidebar. It shows suggested budget steps, and lists changes already made through the Marketing API. Facebook Login: not used. Bridge has no Facebook Login. People sign in with email and password or Google, and Bridge connects to Meta with a system user access token set up by our team. We do not request email, public profile or other user data permissions from Facebook users. Meta APIs used: the Marketing API (campaigns, ad sets, ads, creatives, images, videos and insights), Pages listing to choose the Page an ad runs from, and the Conversions API to send conversion events to our own datasets.
Is Facebook Login integrated on this platform? No.
Membership is by invitation only, so here is a read-only test account: Email: [test email] Password: [test password]
Not applicable. Bridge is a website, so there is nothing to download or pay for.
Not applicable. Access is not restricted by location.
Sign in once with the test login to make sure it works, then submit for review. Meta says this typically takes under a week, often 2–3 days. Watch Alert Inbox in the sidebar, where Meta raises questions.
Meta may ask for Business Verification. It needs a registered company, and an official phone number, postal address or verified domain. You start it in the Security Centre.
To keep Full access, Meta asks for reviews each year, including a Data Use Checkup, so keep the test login active and look at Required actions in the dashboard from time to time.
Meta’s own pages, if a screen differs from this guide:
Campaign template
How a Meta campaign for a Bridge app is built — taken from a live subscription app’s account, where it has been run and tuned across 16 campaigns. Campaigns Claude creates are built exactly like this; build by hand the same way.
| Structure | One campaign per language market and platform, one ad set in it, several ads (about 7 in the account it is taken from). |
| Names | English Campaign - Bridge iOS · English Ad Set - Bridge iOS · ads Concept_LANG_Bridge_iOS, e.g. Reminders_2_EN_Bridge_iOS. |
| Campaign | Sales. Budget on the campaign (Advantage campaign budget), daily, highest volume. Start a market around $25 a day and raise the ones that convert — the account’s best market runs at $150. |
| Conversion | Website, Purchase on the app's own dataset. |
| Audience | The market's countries (people living in or recently there), ages 18–65, Advantage+ audience on. A language filter where countries share languages (French, German). |
| Devices | Mobile, only the OS the link leads to: an iOS campaign for the iOS link, an Android one for the Android link. |
| Placements | Facebook, Instagram, WhatsApp, Threads. Not Audience Network. |
| Attribution | 7-day click, 1-day view, 1-day engaged view. |
| Ads | Single image, Download button, the Instagram account as identity, the app's Bridge link with its macros. Text is optional: the account’s best ad has no primary text or headline — the image carries the message. |
| Enhancements | On: CTA, brightness and contrast, image templates, touch-ups, comments, text optimisations. Off: animation, translation, media text replacement, site and product extensions. |
Markets. English: US, UK, Canada, Australia, New Zealand, Ireland · Spanish SA: Mexico and 15 more Latin American countries · Spain · Brazilian: Brazil, Portugal · French: France, Belgium, Switzerland (French speakers) · German: Germany, Austria, Switzerland, Luxembourg (German speakers) · Italian · Polish. Any market reaching the EU — Ireland included — needs the ad account's DSA payer and beneficiary.
When it goes wrong
Every entry below is a failure this system actually produced, with the symptom that appeared instead of an error.
| What you see | What it is |
|---|---|
| A setting changed in SQL has no effect | App config is cached in KV. The console invalidates it;
direct SQL does not. Worse, capi_enabled = 0 marks
events skipped terminally — they are
never retried when it is switched back on. Change settings in
the console. |
| Test events never appear | The test event code belongs to a different dataset. Meta
accepts a foreign code, returns events_received: 1,
and the event goes nowhere. There is no error to find. Copy the
code from this dataset's Test Events tab. |
| Sandbox purchases produce nothing | Deliberate. Sandbox events are dropped unless the app has a test event code, so they can never reach the live pixel and train the optimiser on money that does not exist. |
error_subcode 2804043 |
action_source: "app" without an
app_data.extinfo block. Unattributed events are
sent as "other" for exactly this reason. |
error_subcode 2804004 |
Event timestamp in the future, from an accelerated sandbox clock. Timestamps are clamped to now. |
error_subcode 2804050 |
Empty user_data — nothing for Meta to
match on. Happens if the click row disappeared before its event
was delivered. These are now dropped rather than retried. |
| Match rate looks terrible | Probably not a matcher problem. The denominator counts only installs that had a click; organic installs and Google Play's pre-launch test devices appear under No click available. Check that first. |
Lots of relay_unmatched |
iCloud Private Relay or a VPN. No usable IP signal, so those are matched on a 2-hour window at a much higher threshold and mostly go unmatched by design. |
A purchase is unattributed |
Either genuinely organic, or the install never matched. The unattributed share is the honest read on revenue the bridge failed to connect. |
Reference
- Event map — the default reports every initial
conversion as
Purchaseand suppressesSubscribe. Meta optimises against one event, and splitting a single commercial moment three ways means each needs its own ~50 weekly conversions to leave the learning phase. - Trials report as Purchase with value 0. Fine when
optimising for purchase count; if you ever switch to
value optimisation, revisit it — the revenue arrives later
on
TrialConversion. - Renewals are never optimisable and stop after 12 months, so ROAS cannot inflate without bound.
- One-off purchases (lifetime unlocks, coins, credits)
report as
Purchasethe first time a person buys that product. Buying the same product again reports asRepeatPurchase: revenue, never optimisable, and like renewals it stops after 12 months. - Debug and production are separate registrations with separate keys. A debug build cannot reach production data.
Plans — connect the stores
Plans shows each app’s subscriptions as the App Store, Google Play and RevenueCat have them: the price in the United States, the store’s state, the offerings each is in, and how many people pay for it now, one app at a time. Price tests start from it and split new users between offerings, and the Bridge connector makes new products in both stores and in RevenueCat. Three connections, each made once for the workspace.
The key on Setup → RevenueCat subscriptions. In RevenueCat, API keys → … → Edit, open each group with its arrow and set:
- Customer information: Customers Configuration on Read & write (a test puts each new user on its offering); Subscriptions, Purchases and Invoices on Read only.
- Project configuration: Products, Packages, Offerings and Entitlements on Read & write (new products made through the Bridge connector are registered there, given the app’s entitlement and put in an offering); everything else on Read only.
- Charts & metrics, Audiences and IAM: Read only.
Then Submit. The key stays the same: nothing to paste again. Each app’s RevenueCat project has its own key on Setup; an app without one shows only what the stores say, and Plans says so.
In App Store Connect → Users and Access → Integrations → App Store Connect API,
Team Keys → +. Name it Bridge with the App Manager role, which reading and changing
subscriptions and their prices needs, and Generate. Download the .p8 file (Apple lets you
download it once), and copy the Issuer ID and the key’s Key ID. Paste all three on
Setup → App stores: Bridge checks them with Apple, keeps the key encrypted and
never shows it again. Delete the file then. One key covers every app in that App Store Connect team, and Setup lists
each iOS app it can read. Bridge keeps one key for the workspace, so an app in another Apple team cannot be read yet.
Enable the Google Play Android Developer API in the Cloud project of the service account Bridge uses for Google Ads. In Play Console → Users and permissions, Invite new users with its email (on Setup → Google Ads API), or open it if it is there already. Under Account permissions, tick View app information and download bulk reports (read-only) and Manage store presence, which covers subscriptions and their prices: that reaches every app in the developer account, ones added later too. To give it some apps only, use App permissions instead, app by app. Play adds View app quality information and Manage policy declarations on its own. To release Android builds through Bridge as well, tick Release apps to testing tracks, and for production Release to production, exclude devices, and use Play App Signing: without them Google takes the upload but refuses the release. Nothing else is needed, and no financial data. Then on Setup → App stores, Use the Google Ads service account, and Check again after any change. Apps in another developer account need the same there. Without Google Ads connected, paste a service account key of its own instead.
What a price test may change
- It reuses before it makes. A product nobody holds — in no live offering, with no one paying for it and nothing bought on it for 90 days — is repriced for a test instead of a new one being made. Plans marks each one Free to reuse or says why not. A product people hold is never repriced: its price is theirs.
- It names new products the app’s way. Plans shows the pattern read off each store’s own ids,
such as
myapp_{annual|monthly}_{sub|pro|premium}on the App Store andmyapp_premium:myapp-{annual|monthly}-{…}on Google Play, and which word means which length. A new product follows it, and never takes an id the store has seen before: Apple never lets a product id be used twice, even after it is deleted.
Plans — run a price test
A price test splits new users between the current offering in RevenueCat (the control) and another offering of yours at other prices (the variant), and tells you which brings more money per user. It uses offerings and products you have already: nothing is made in the App Store or Google Play.
On Plans, pick the app, then under Price tests the offering to test. Bridge shows each package’s price now and in the test, per store, and Phones: iPhone and Android, or one of them. When the offering changes only one store’s prices, that store’s phones are picked for you, so users on the other store, who would see the same prices either way, do not water the result down. An offering with the same prices on the phones picked is refused. A price that moves by more than 25% asks you to be sure. Start the test. One test at a time per app.
Or Ask the AI for a test: it reads the app’s offerings and their prices per store, what new users bought and paid in the last 30 days on each platform, and the app’s earlier tests, and proposes the one test most likely to bring more per user: the offering, the phones, why, what would show it worked, and what to watch. It proposes only offerings you have, within 25% of today’s prices, and nothing starts until you Approve and start it; Dismiss it, or ask again later. Each ask is one AI analysis, billed like a plan on Automation.
Every new install on the phones picked, from its first launch after the start, for up to six weeks: half get the variant, half keep the current offering, by a coin fixed to the install. Bridge puts each one on their group’s offering in RevenueCat, and the app, which shows RevenueCat’s current offering, shows them that one. People who had the app before are not touched.
The app’s first paywall can load a moment before Bridge has put a new user on their offering; those few see the current one on that screen and the variant after. They are still counted in their group, so a real difference reads a little smaller, never larger.
Proceeds per user: what each user paid in their first 30 days, after the stores’ fees and tax, in USD, trials that paid nothing included. Bridge reads both groups hourly: users, trials, payers and proceeds per user, so far and 30 days on, and the same split by where users came from (Google ads, ads through Bridge’s links, organic). A group wins when it is 95% likely to bring more per user, with at least 200 users 30 days on in each group. Until then Plans says how likely the variant is to be better.
End stops new users joining; the ones in it keep their offering while their 30 days run. When you decide, Roll out the variant makes it the current offering in RevenueCat for everyone, or Keep the current offering leaves it. Either way Bridge lifts the test’s overrides, a few hundred an hour within RevenueCat’s limits, and everyone sees the current offering again.
Testing a price no offering has yet needs new products in the stores. The AI lists such tests apart, under Tests that need new prices. Bridge makes the products through the Bridge connector (create_store_products, then add_revenuecat_products): it shows what it would make before making anything. Put them in an offering in RevenueCat and it can be tested.
Google Ads — connect your account
Lets Bridge work in your own Google Ads account: read each campaign’s spend and results every hour, so Campaigns, Analytics and Store payouts count Google spend beside Meta’s; plan and create App and Search campaigns (always paused), change their daily budgets, keywords and ad text and switch them on and off, each shown to you first. One account per workspace, connected once in Setup → Google Ads API.
Bridge signs in as a service account: a Google identity that belongs to your Cloud project, not to a person, added to Google Ads like any other user. Nothing to renew, and it reaches only the Google Ads accounts you add it to. Since September 2026 the Google Ads API needs no developer token; access belongs to the Cloud project instead. Google does not charge for the API, and the project needs no billing.
What you need first
- A Google account that is an admin of the Google Ads account, to add a user.
- A Google account that can create a project in Google Cloud Console. It can be the same one.
In Google Cloud Console, New project, named after your company or this workspace, not one app: for instance mycompany-bridge. The service account’s email carries the project’s id (bridge-ads@mycompany-bridge.iam.gserviceaccount.com), and that one email reaches every Google Ads account of the workspace, whichever app it is for. A project you already have works too: Google grants the API’s access level per project, so keep using the same one.
With that project selected, open the Google Ads API in the API Library and click Enable. The project then has Test access, which reaches test accounts only.
On the API’s Overview page in Cloud Console, open Upgrade access level and click Apply for access: the next level is Explorer, up to 2,880 operations a day on real accounts, of which Bridge’s hourly reads use a few dozen. There is no form: the page says the application is under review, and a few minutes later it shows Current access level: Explorer. Explorer needs no brand verification, and covers everything Bridge does with your campaigns. Basic raises the daily limit and has Keyword Planner answer from your own project; Google grants it only once the project’s brand is verified: step F, whenever you want it. How far each level goes: how much Bridge may ask of Google.
Enable the Data Manager API in the same project as well. Bridge sends the trials and first payments of Search clicks through it, because Google closed Google Ads’ own conversion upload to newcomers in June 2026. It has no access levels or developer token; until it is on, Setup says the conversions could not be sent.
In the same project, IAM & Admin → Service accounts → Create service account. Name it bridge-ads and give it these roles: Service Usage Consumer, which the Data Manager API asks for, and BigQuery Data Viewer and BigQuery Job User, with which it reads Google Analytics’ export for App campaigns in RevenueCat, when the export is in this project. Google Ads access is given in Google Ads, in step D. Done. An account made before this: IAM → Grant access, its email, the same roles.
Open it, then Keys → Add key → Create new key → JSON → Create.
A .json file downloads. It works like a password: keep it out of email and chat,
and delete it once it is in Bridge. Its client_email, ending in
.iam.gserviceaccount.com, is the address for step D.
In Google Ads, Admin (the gear) → Access and security → Users → +. Paste the service account’s email and choose Standard access, which changing budgets and switching campaigns needs; Read only is enough if Bridge should only read. For a service account the button says Add account, and it is added at once, with no invitation to accept: Google confirms it in a message at the bottom. It does not appear in the Users list, so do not add it twice.
Use the email exactly as client_email in the key file has it. One service account can serve
Google Ads and Google Play alike; if Google Play is connected already, it is the email on
Setup → App stores.
“This email address isn’t in an allowed domain”: the account lets in only its own email
domains, and a service account’s domain is its Cloud project’s, such as
my-project.iam.gserviceaccount.com. An admin adds that domain under Admin → Access and
security → Security → Allowed domains, then adds the service account again.
“This request needs to be approved by another admin”: Google asks a second admin to approve Standard access. Request approval; another admin of the account, or of a manager account above it, gets an email and approves within 20 days. Bridge can connect once it is approved.
In Setup → Google Ads API, paste the whole file, and the account’s customer ID: the ten digits at the top of Google Ads, such as 123-456-7890. Give a manager ID only when you reach the account through a manager account (MCC) and added the service account there. Connect: Bridge asks Google whether it can read the account, keeps the key encrypted, never shows it again, and reads the last 30 days at once.
Basic raises the project’s limit from 2,880 to 15,000 operations a day, and has Keyword Planner’s figures (searches a month, competition, bids) come from your own project. Without it nothing is missing: Bridge gets those figures through its own project, within a daily allowance, without touching your account.
Google grants Basic only to a project whose brand is verified. Until it is, Upgrade access level on the API’s Overview page says Brand Verification needs to be completed on your account to be approved for Basic access, and an application is refused.
Brand verification, then the application: step by step
As in Google’s brand verification guide, all in the Google Cloud project Bridge is connected with:
- In Cloud Console, open Google Auth Platform (APIs & Services → OAuth consent screen). If it is not set up yet, Get started: an app name, a support email, External audience and a contact email, then Create. Bridge does not sign anyone in through it; Google asks for it for Basic all the same.
- Audience: the user type must be External (Make external if it says Internal) and the publishing status In production (Publish app, then Confirm). Other Google pages say an Internal or Testing app needs no verification; for Google Ads API Basic access it does.
- Branding: your app or company name, a logo that is your own brand’s (120×120 pixels), your home page, privacy policy and terms of service on a domain you own, and that domain under Authorized domains. Save.
- Verify the domain as yours in Google Search Console, with the Google account that owns the Cloud project.
- Back on Branding, under Verification status at the top right, Verify branding. Often it takes minutes; when Google’s team looks at it, a few days. If it lists issues, fix them, wait the day it asks for, then View issues → I have fixed the issues. If it flags your logo wrongly, choose I believe the issues found are incorrect and say where the logo is used.
- Once verified, Publish branding. Branding then shows the project as brand verified.
- On the Google Ads API’s Overview page, Upgrade access level: the next level says Basic, with no brand warning. Apply for access: a short form on what the tool does. Google reviews it, usually within a few business days, and emails the outcome.
Once granted, set the level to Basic on Setup → Google Ads API (Bridge also learns it from Google’s answers). Keyword figures then come from your own project, with no daily allowance, within the hour.
More than one Google Ads account
One account serves every app by default. When some apps’ campaigns are in another Google Ads account, add it on Setup → Google Ads API, Add another Google Ads account: first, in that account, Admin → Access and security → Users → +, add the same service account’s email with Standard access (no new key); then give its customer ID and tick the apps it is for. Bridge reads every account’s campaigns hourly and files each under the app it promotes; the apps ticked have their new campaigns and keyword plans made in that account, and every other app keeps the first one. Budgets, keywords and ad text are changed in whichever account a campaign is in.
An account may be in another currency. Each campaign’s figures stay in its own account’s currency; Automation’s goals stay in the first account’s, and another currency’s budgets are held to them at the day’s rate (a step is refused when no rate is to be had).
What Bridge does with it
- Spend and results, hourly. Each campaign’s spend, impressions, clicks and Google’s own conversions, in the account’s currency and time zone; the whole last month again every six hours, as Google revises recent days.
- Filed under your apps. An App campaign is filed under the Bridge app it promotes, by its App Store number or Play package; a Search campaign by the Bridge link in its tracking template. Other campaigns are listed on Setup but not counted under an app.
- From Claude, with the Claude connector:
google_ads_campaignsreads them,google_campaign_detailsone in full (its keywords, ads and Google’s rating of each line),google_automation_planthe AI’s latest plan and what it changed;create_google_search_campaignmakes a Search campaign on chosen keywords (see run Search campaigns);create_google_app_campaignmakes an App campaign for one store, with its budget, bidding goal, countries, languages and an ad of headlines, descriptions, images (links or Figma frames) and YouTube videos, always paused;set_google_campaign_budgetandset_google_campaign_statuschange a daily budget or switch campaigns on and off. Each shows the change first, Google checks it without making it, and nothing changes until you say yes. A new campaign is made whole or not at all. Only a manager of the app can change its campaigns. A budget shared between campaigns is left to Google Ads. - Automation. Google has its own AI plan, goals and autopilot, beside Meta’s: switch Automation to Google Ads. The autopilot changes budgets, a Search campaign’s keywords and the ad lines Google rates Low, each within your goals and each kind on its own switch. Plan a new campaign there drafts a Search campaign for you to check and create.
- Keywords. The Keywords page: what your Search campaigns target and block, the searches that showed your ads, and a check of any keyword, with Add and Block.
On Android, an App campaign’s click reaches Google Play with Google’s click id in the install referrer, usually with the campaign’s id beside it, which Bridge’s SDK reads. So Bridge credits each Google campaign with the installs that name it, and those with a click id alone when it is the app’s only Google campaign that spent that week, and everything those users pay, and sets that against its spend, the way it does for Meta’s link clicks. Facebook’s own referrer can carry a placeholder click id; Bridge counts that as Meta’s. Google’s own conversion counts stay beside it, and can run a little higher. iOS gives no referrer: an iOS App campaign shows its spend and Google’s counts only. For proceeds per Google campaign inside RevenueCat, see the GA4 route below.
When Google refuses
| Bridge says | Do this |
|---|---|
| The Google Ads API is not enabled | Step B, in the project the service account belongs to. |
| Test access only | Step B’s Apply for access. Until Google grants Explorer, real accounts refuse the project. |
| The service account cannot open the account | Step D, with the email shown; or the customer ID is another account’s. Through a manager account, give its ID too. |
| That is a manager account | Give the ID of the account the campaigns are in, and the manager’s as the manager ID. |
| Google refused the service account’s key | The key was deleted in Cloud Console, or the file was cut short. Create a new key (step C) and connect again. |
| This method is not allowed for use with explorer access | Keyword Planner only: the project needs Basic access, step F. Everything else works. |
| Auto-tagging is off in this Google Ads account | Google Ads → Admin → Account settings → Auto-tagging: on. Without it a Search click carries no Google click id, so Bridge can neither tell it from a check of the link nor send its trials back to Google. |
Google Ads — how much Bridge may ask of Google
Google limits what a Google Cloud project may ask of the Google Ads API in any 24 hours, by its access level. Your project is yours, so the limit is yours alone: no other Bridge client uses it. Google counts operations: a report request is one however many rows it returns (further pages are free), a conversion upload is one for up to 2,000 conversions, and each change (a budget, a keyword, an ad line) is one. So an app’s size hardly matters: even one with thousands of installs a day from Google Ads takes Bridge a few hundred operations a day.
| Access level | A day, on live accounts | How you get it |
|---|---|---|
| Explorer | 2,880 | Apply on the Google Ads API’s Overview page; usually granted at once (connect your account). |
| Basic | 15,000 | The project’s brand verified, then apply. Keyword Planner’s figures (searches a month) need it: below it, Bridge gets them through its own project, within a daily allowance, without touching your account (step F). |
| Standard | No limit | From Basic, an application Google reviews by hand, in about 10 business days (below). |
What Bridge uses, and when it asks you
- About 85 a day to read spend and results hourly, up to 24 to upload conversions, a few hundred at most to look up the keywords of Android installs, and some tens for AI plans, the autopilot and the Keywords page; a new campaign takes one per thing it makes, around a hundred.
- Setup → Google Ads API shows Bridge’s own operations in the last 24 hours against your level’s limit. Set the level there as the API’s Overview page shows it; Bridge also learns it from Google’s answers. Until it knows, it goes by Explorer’s 2,880.
- From 80% of the limit, or as soon as Google refuses a request for it, the console asks an admin, once a day, to apply for the next level.
- Bridge sees only its own requests. Another tool using the same Cloud project counts towards the same limit: give Bridge a project of its own.
- At the limit, Google refuses until enough of the last 24 hours have passed. Hourly reads and uploads simply wait and catch up; a change you make then fails with Google’s reason, and can be made again later.
Apply for Standard access
- In Cloud Console → Google Ads API → Overview, check that Current access level says Basic. Standard is only open from Basic.
- Open Upgrade access level; the next level should say Standard. Click Start application.
- Google audits the application by hand. Be ready to say what the tool does and who uses it: Bridge, an attribution and campaign management tool our team uses for our own apps’ Google Ads accounts: hourly reporting, budget and status changes, Search keywords and ad text, campaign creation, offline conversion uploads and Keyword Planner research. Name your daily operations (from Setup) and why they grow. If people outside your company use the tool, Google asks for a demo sign-in, and a tool that manages campaigns must meet Google’s Required Minimum Functionality.
- Google’s compliance team writes to the project’s contact email: answer promptly, as the review waits on you. It usually takes about 10 business days, and Google emails the outcome.
- Standard comes with a permissible use. Bridge needs ad creation and management (budgets, keywords, ad text, new campaigns), reporting and researching keywords; ask for all three.
- Once granted, set the level to Standard on Setup. Nothing else changes in Bridge.
Google Ads — run App campaigns
How Google campaigns are set up to bring in paying users, start to finish. An App campaign is the one kind of Google campaign made for installs: a single campaign shows your ads in Google Play, Google Search, YouTube, Discover and the Display network. There are no keywords to choose. Google matches searches, in Play and on Search, from your store listing and the text you give it, and learns from the in-app action you ask for which users to show the ads to. So the levers are the store listing, the assets, the action it bids on, the countries and the budget.
The account connected to Bridge (connect your account), a payment method in Billing, and advertiser verification done (Billing → Advertiser verification). Until it is, Google serves a new advertiser’s ads to few people: a first campaign can spend nothing at all until it is.
- The app gives RevenueCat its Firebase app instance id: the code under Part 1 below. RevenueCat → Integrations → Firebase: each platform’s Firebase App ID and Measurement Protocol API secret (GA4 → Admin → Data streams), revenue after store commission and taxes, in USD, sandbox events off.
- For iOS, Firebase → Project settings → the iOS app → App Store ID. Without it Google Ads does not offer the iOS stream at all.
- GA4 → Admin → Product links → Google Ads links → Link, and pick the account. Its defaults are fine.
- GA4 → Admin → Events: star
rc_trial_start;purchaseis a key event already. Google Ads imports key events only. Unstar any event it starred that the app does not send. - Google Ads → Goals → Conversions → Create conversion action →
Conversions on an app (untick phone calls) → Edit data sources, tick
Google Analytics → Save and continue → Create multiple conversion actions from a
linked account → the GA4 property. Tick
purchase(category Purchase) and the starredrc_trial_start(Signup), for each app stream. Value from Firebase, every conversion, 90-day windows: the defaults are right. - Make purchases the account’s goal: Google Ads → Goals → Summary, then on Purchase → Edit goal → Account default → turn on Make this an account-default goal → Save. Do the opposite on Download and Sign-up, so Purchase is the only goal marked Account-default. Both keep recording. Google’s Conversions column, cost per conversion and overview then count purchases rather than installs, and any Search or Demand Gen campaign you add aims at them by default. An App campaign bids on the actions picked in its own settings (step C), so this changes no running campaign and restarts no learning.
in_app_purchase or
app_store_subscription_*: they are the same sales as RevenueCat’s
purchase, and importing both counts each one twice.New actions say Awaiting conversions for a day or two. Setup → Google Ads API lists what the account can bid on, and warns when nothing in-app is there yet.
From Claude, with create_google_app_campaign: say the app, platform, market,
budget and that it should aim at purchases, and give the headlines, descriptions, images and
videos. It shows the plan, Google checks it, and it is made paused. Or in Google Ads,
+ Create → Campaign → App promotion → App installs, and the app. Either
way:
- One campaign per market and platform, named like Meta’s: English MyApp – Android Purchases, its ad group English Ad Group – Android Purchases.
- Countries: the same as the Meta market’s, for English US, GB, CA, AU, NZ and IE. Under Location options, choose Presence, people in those countries, rather than people anywhere interested in them. Language: the market’s.
- Bidding: In-app actions, then
purchasewhen the platform makes ten or more purchases a day from every source; otherwiserc_trial_start, which comes sooner and more often. Leave the target cost empty for the first weeks. Once purchases have values for a few weeks, In-app action value aims at revenue rather than count. - Budget: where the Meta market runs. Google may spend up to twice the daily budget on a day, never more than 30.4 days’ worth in a month.
- Assets: 5 headlines of up to 30 characters, 5 descriptions of up to 90, up to 20 images (1200×628, 1200×1200 and 1200×1500), and videos on YouTube, unlisted is fine: one or more each of landscape 16:9, portrait 9:16 and square 1:1, 10 to 30 seconds. Without videos Google makes its own from the images, and ad strength stays Poor.
With set_google_campaign_status, which says what it will spend first, or in Google
Ads. An App campaign learns for about a week, and again after a big change: no changes in that
time, then budget steps of 20% at most with three days between them, as the
Automation tab’s Google plan and autopilot keep to.
Eligible (Limited) for a sensitive category in personalized advertising is expected for an app in one of Google’s sensitive categories. Google does not use personal interests to target those ads, so they serve without them. There is nothing to fix or appeal.
- Campaigns, with the Google Ads source: spend and clicks are Google’s; installs, purchases, revenue and ROAS are Bridge’s own, from the Android installs whose Play referrer names the campaign or carries Google’s click id. With two Google campaigns for one app running at once, only the ones that name their campaign are credited. iOS gives no referrer.
- Automation → Google Ads: Google’s installs or in-app actions, whichever the campaign bids on, beside what Bridge saw its users pay, an AI plan, your goals and the autopilot. Google’s in-app actions arrive from Firebase hours to a day late. Each plan also reads the year (the app’s new customers by week, Google’s monthly searches for each Search campaign’s keywords, your occasions) and raises budgets before a season the app had last year, each step with the day it ends; Bridge sets each back that day, unless you changed it since.
- When a campaign’s goal changes, rename it and its ad group to match. Bridge reads names hourly.
Google Ads — run Search campaigns
A Search campaign shows a text ad on Google Search to people looking for an app like yours, on the keywords you choose. Unlike an App campaign, you pick the searches, so you can see which ones bring paying users. Bridge sees them on Android: Google Play passes Google’s click id to your app, and Bridge looks up the campaign and keyword it came from. An iPhone install from a Search ad carries nothing from the ad, and Google allows no Bridge link in Search ads (it has not certified Bridge as a click tracker), so plan Search campaigns for Android first.
The account connected (connect your account), a payment method and advertiser verification (run App campaigns, step A), and auto-tagging on, as it is by default: Google Ads → Admin → Account settings. Bridge checks it before it makes a campaign.
Automation → Google Ads → Plan a new campaign. Pick the language (your app’s own languages, from its App Store listing, come first) and the phones: Android only (the default), iPhone and Android, or iPhone only. Budget, countries and notes are optional. The AI drafts, from your store listing in that language, the searches your campaigns have shown for and your goals:
- Keywords, each with why, phrase or exact match, never broad;
- the searches it left out, each with why, with Target it to take one in;
- blocked searches (negatives), such as PDFs, shopping or unrelated topics;
- 15 headlines of up to 30 characters and 4 descriptions of up to 90, saying only what your listing says the app does;
- the budget, countries and bidding: as many clicks as the budget buys, for a first campaign.
Google checks the whole campaign without making it. Change anything, Check with Google
again, then Create paused. Claude can do the same with create_google_search_campaign.
Each draft is one AI call, billed like a plan.
- Google Search only, phones only (no one installs from a computer), people in the countries rather than interested in them.
- For iPhone and Android: one ad to your Google Play page, with an App Store link that Google shows on iPhones. Google sends every click on an ad to that ad’s own page, so an iPhone user who taps the main ad sees the Google Play page; the App Store link sits under it.
- For one platform: Google shows Search ads on every phone and cannot be told otherwise, so the ad
goes to that platform’s store only, a headline saying so (MyApp for Android) is pinned first so
people on the other platform skip it, and their searches (iphone, ios… or
android, google play…) are blocked. Bridge files it under that platform’s
registration. Claude moves a running campaign between them with
set_google_search_platforms, shown to you first. - No tracking template. Google disapproves Search ads that go through a click tracker it has not
certified, and it has not certified Bridge (it takes applications again from September to November 2027).
Bridge files the campaign under the app whose store page its ads go to. A campaign made earlier through
Bridge’s link is disapproved until the link comes off: Claude takes it off with
set_google_search_platforms, shown to you first. - Conversions are your account’s own, from Firebase or Google Play (Google Ads → Goals → Conversions): the check with Google names the ones it will count for the app, or says there are none. Bridge sends none of its own, as Google would count them twice. Link the app’s Firebase before bidding on conversions.
The Keywords page shows what the campaign targets and blocks, the searches that showed your ads, and a check of any keyword: how often it is searched, whether you target it and which blocked search would stop it, with Add and Block. Searches a month need Google’s Basic API access (connect your account, step F); until then, Google Ads → Tools → Keyword Planner gives them by hand.
In Google Ads, or ask Claude (set_google_campaign_status, which shows what it will
spend first). Google may spend up to twice the daily budget on a day. In
Automation → Google Ads → Goals, set Target cost of a trial ·
Search campaigns and what the autopilot may do: change Search budgets, change keywords (pause,
retest, add, block, at most 15 a week), and replace the headlines and descriptions Google rates
Low (at most 3 an ad in two weeks, each undoable). It leaves keywords alone until a campaign
has had 100 clicks or a week of spend, and Google rates new lines after a week or two. Once the
campaign brings about 30 trials a month, bidding on conversions can replace bidding on clicks.
- Campaigns, Google Ads source: Google’s spend and clicks, Bridge’s installs, trials, revenue and ROAS.
- Keywords: the same per keyword, and the searches behind them.
- In RevenueCat, each subscriber from a Search ad carries Bridge’s
bt_networkgoogle,bt_campaign,bt_keywordandbt_match_type; and, where RevenueCat’s own fields are turned on for the app (the owner of Bridge does this),$mediaSourceGoogle Search,$campaign,$adGroupand$keyword. Both need Bridge SDK 0.6.1 on iOS or 0.3.1 on Android, or later. - Android installs: each carries Google’s click id from Google Play. Bridge asks Google’s click report which campaign, ad group and keyword that click came from (Google can take up to a day to have it; Bridge asks a few times over three days, all of a day’s clicks in one request) and writes the same fields to the RevenueCat customer, and counts the install under its keyword. That needs the RevenueCat key on Setup to write customers: in RevenueCat, API keys → … → Edit, open Customer information permissions and set Customers Configuration to Read & write, with Subscriptions, Purchases and Invoices left on Read only, then Submit. An iPhone install from a Search ad carries nothing from the ad, so it gets nothing. Setup → Google Ads API shows how many were found and written.
Google Ads — App campaigns into RevenueCat (GA4)
Your values
Two results, set up independently. Conversions to Google Ads, so
Google's bidding learns from real subscription revenue. And campaign
and ad group on each RevenueCat subscriber — media source
Google UAC, $campaign, $adGroup —
so revenue can be split by Google campaign inside RevenueCat.
How it works
Google does not give user-level App campaign attribution to anyone but
its certified attribution partners: the click id is withheld for App
Campaigns for Installs, and iOS never exposes one at all. Android’s Play
referrer usually names the campaign, which the SDK writes to RevenueCat
(Google UAC and $campaign), but never the ad group, and some
App campaign clicks (those on Google Search) come with no campaign at all. Google
Analytics knows which campaign and ad group acquired each install, and its
BigQuery export records it per user. Bridge reads that, matches each user
to their RevenueCat customer through the Firebase app instance id, and
writes the result back to RevenueCat.
Before you start
- Firebase Analytics is in the app, in the version people have installed: the
FirebaseAnalyticspackage on iOS,firebase-analyticson Android. Other Firebase products (Auth, Firestore, Messaging, Remote Config) send Google Analytics nothing. The app’sGoogleService-Info.plistandgoogle-services.jsoncome from the same Firebase project as the export. To see it working: Google Analytics → Reports → Realtime shows the app’s users. - Google Analytics is enabled on the Firebase project (Firebase → Project settings → Integrations), and that GA4 property is linked to Google Ads (GA4 → Admin → Google Ads links).
- The app is registered in Bridge (Setup → Apps → Add an app) and its RevenueCat webhook points at Bridge.
Part 1 — conversions to Google Ads
1. Give RevenueCat the Firebase app instance id.
One line, right after Purchases.configure. It must come from the
Analytics package — RevenueCat warns that a wrong id silently stops
events reaching GA4.
// iOS
if let id = Analytics.appInstanceID() {
Purchases.shared.attribution.setFirebaseAppInstanceID(id)
}
// Android
FirebaseAnalytics.getInstance(context).appInstanceId.addOnSuccessListener { id ->
if (id != null) Purchases.sharedInstance.setFirebaseAppInstanceID(id)
}
2. RevenueCat → Integrations → Firebase. Add the Firebase App ID and a Measurement Protocol API secret, both from GA4 → Admin → Data streams. iOS and Android are separate streams, so each needs its own. Choose net revenue if Google's ROAS should reflect what actually lands after the store's cut, gross to match store reports — and do not change it mid-campaign.
3. Key events, the Google Ads link and the import.
As in run App campaigns, step B: import
purchase and rc_trial_start for each stream, never
Firebase's own in_app_purchase, and let each campaign’s bidding
choose which of them it aims at.
Part 2 — campaign and ad group in RevenueCat
Needs Part 1, step 1 (the Firebase id on RevenueCat), the Google Ads API connected (connect your account), and the RevenueCat key on Setup able to write customers (Customers Configuration: Read & write).
1. Firebase → Project settings → Integrations → BigQuery → Link.
Turn on Google Analytics, pick a region (US is fine), tick Daily (streaming
is not needed), and leave advertising identifiers off. It creates a dataset named
analytics_ followed by the GA4 property id, the first table a day later.
The export is not retroactive: installs from before the link have no data. Google makes the dataset
with the first export, so until the app’s users send events it does not exist.
2. Choose what is exported. In GA4 → Admin → BigQuery links →
the link → Configure data streams and events: tick every stream of the app that is live,
iOS and Android: the ones Firebase lists under Project settings → Integrations → Google
Analytics. A stream not listed there belongs to no Firebase app and never gets events.
Bridge reads one event, first_open: it must not be under
Events to exclude, or nothing reaches Bridge. The list to exclude from shows only events the
ticked streams sent lately: No events available, or a daily event volume of 0, means the app
sends Google Analytics nothing yet (see Before you start).
Without a billing account the project is on BigQuery’s free sandbox: 10 GB of storage
for its whole life, after which the export stops for good, even if data is deleted. Then keep it
small: under Events to exclude, add the app’s other events by name (such as session_start,
screen_view and user_engagement), never first_open, and the free tier lasts years. With a billing
account the daily export is free, storage costs about $0.02 per GB a month after the
first 10 GB, queries $6.25 per TB after the first 1 TB a month: with tables kept 70 days
(the dataset’s default table expiration), cents a month.
3. Google Cloud → IAM, in the export’s project → Grant access. Add the Google Ads API’s service account (Setup shows its email) with BigQuery Data Viewer and BigQuery Job User. No new key: Bridge already holds that account’s. Already done when the service account was made with these roles in the project that holds the export; an export in another project, such as the app’s Firebase project, needs them given there.
4. Setup → GA4 → RevenueCat. Choose
the app, enter the project id and the dataset, tick Run nightly, save. The project id is the
Firebase project’s (Project settings → General); the dataset is analytics_
followed by the GA4 property id, which Firebase shows as Property ID under Project settings →
Integrations → Google Analytics (also Google Analytics → Admin → Property details), for
example analytics_123456789. It is listed under the project in
BigQuery
once the first day is exported; it can be entered before.
5. Check now. It runs a dry run: nothing is written, and it reports how many users it would write and examples of the campaign and ad group. GA4's export lands about a day late, so a user who installed today appears tomorrow. Before the first export it says the dataset was not found: expected, and it clears by itself.
6. Switch to Write once the examples look right, and save. From then on it writes nightly.
When it does not work
| What you see | What it means |
|---|---|
| No events available when choosing events to exclude, or a daily event volume of 0 | The ticked streams send nothing: Firebase Analytics is not in the build people have, or its config file is from another Firebase project, or the stream is not one of the Firebase apps’. See Before you start and step 2. |
| Check now: Not found: Dataset … analytics_… | No export yet (above), or the dataset name is not analytics_ with the property id Firebase shows. It clears by itself the day after the first events. |
| Check now: bigquery.jobs.create or Access Denied | The service account lacks BigQuery Job User, or BigQuery Data Viewer, in the export’s project: step 3. |
| The dry run finds no users | first_open is excluded (step 2); the installs came before the link; or the app does not give RevenueCat the Firebase id (Part 1, step 1), so no customer can be matched. |
What it will and will not claim
- Only users Google Ads acquired. Organic installs, other networks and
GA4's
(not set)are never written as a Google campaign. - On Android it also writes over what the Play referrer gave: the same campaign, now with its ad group, and the campaign where the referrer named none.
- A user Bridge already attributed to a Meta ad click is left alone:
RevenueCat holds one media source per customer, and overwriting it
would re-credit a conversion Meta is being told about. These are
counted as
bridge_attributed. - On iOS, results depend on what GA4 itself attributes. Firebase Analytics brings Google's on-device measurement, which helps, but some iOS installs will remain unattributed. That is a limit of what Google shares, not of this setup.
Web campaigns
For Search or Display campaigns that send people to a page, set the final URL to the app's Bridge link (Setup → Ad links) and add a final URL suffix with Google's ValueTrack parameters:
campaign_id={campaignid}&adset_id={adgroupid}&ad_id={creative}&utm_term={keyword}&utm_source=google
Bridge records them exactly as it records Meta's macros.
Apple Ads
Bridge reads your Apple Ads campaigns on App Store search results every hour: each campaign’s, ad group’s and keyword’s spend, taps and installs, and the searches that showed your ads, filed under the app they promote. From Automation and the connector it changes bids, keywords, blocked searches and budgets, and makes new campaigns, always paused. And for installs of app versions with the Bridge iOS SDK 0.7 or later, Apple tells Bridge which campaign, ad group and keyword brought each user, so a keyword is judged on what its users go on to pay.
Connect your account
On Setup → Apple Ads API, Make Bridge’s key. Bridge makes a key pair: its private half never leaves Bridge, and you paste the public half into Apple Ads. Nothing to download or keep.
In Apple Ads (Advanced), Account Settings → User Management → Invite Users: an Apple Account other than the one you sign in with, with the role API Account Manager (or API Account Read Only, for reading only). It accepts the invitation from the email Apple sends. Bridge reads every ad account the API user can reach.
Signed in to Apple Ads as that API user, Account Settings → API: paste Bridge’s public key and Save. Apple then shows the clientId, teamId and keyId above it: paste them on Setup and Connect. Bridge checks them with Apple at once, then reads the campaigns and their last 90 days, which takes a minute.
If another tool can also change this account’s bids (an agency’s or another bid manager), keep either it or Bridge’s autopilot off: two of them changing the same bids undo each other.
Which keyword brought each user
Apple’s AdServices framework gives each install a token that Apple turns into its campaign, ad group and keyword.
The Bridge iOS SDK 0.7 or later sends it with the install; Bridge asks Apple at once, and again for up to 24 hours, and keeps
the answer with the install as media source Apple Search Ads, with the keyword. Every trial and payment of that
user then counts to the keyword, in Automation, on Keywords and in the autopilot’s cost per trial. Installs of older
versions have Apple’s spend, taps and installs only. Setup → Apple Ads API shows how many installs asked Apple this
week, and what Apple said.
Each iPhone app needs, once:
- The Bridge iOS SDK 0.7.0 or later, in the version people download. An install can be traced to its keyword only in its first day, so installs of an older version are lost to it for good: release the update before the ads run.
- RevenueCat’s Apple Search Ads fields, so the RevenueCat customer shows the campaign and keyword too. In the
app, right after
Purchases.configure:Purchases.shared.attribution.enableAdServicesAttributionTokenCollection(). In RevenueCat, Integrations → Apple AdServices → Add Apple Search Ads integration; for campaign, ad group and keyword names rather than ids, Sign in with Apple under Advanced (a Read Only Apple Ads user is enough). - Users who sign in. When the app gives RevenueCat its own user id (
Purchases.logIn), callBridgeAttribution.shared.identify(rcAppUserId:)with it, as after any login or logout: RevenueCat names the user’s payments by that id from then on, and Bridge moves the keyword with it. - No age or gender targeting on Apple Ads ad groups: since September 2026 Apple answers that their installs came from no ad.
How the answers fit together:
- RevenueCat writes a customer’s source once and never changes it, and its integration asks Apple within a day. A day
after Apple answered Bridge, a customer RevenueCat still left with no source gets Apple’s answer from Bridge:
Apple Search Ads, the campaign, ad group and keyword (Bridge’sbt_fields where the app does not let Bridge write RevenueCat’s own). A customer with a source already is left as it is. - A tap Apple saw beats a click Bridge would credit by likeness alone: Bridge asks Apple before crediting such a click, so nothing wrong reaches RevenueCat. An ad Apple only showed (claim type Impression) does not beat a tap on another network’s ad.
- Builds from Xcode and TestFlight get Apple’s sample answer (campaign 1234567890), which Bridge ignores.
- Trials count as the app reports them: a trial start, or a purchase worth nothing in a trial period where the app sends trial starts to Meta as a purchase.
Keywords
On Keywords, switch to Apple Ads: each keyword the app targets with its bid and Apple’s suggested bid, its spend, taps and installs, its rank and impression share over the last four full weeks, and the searches that showed your ads. Look up a keyword for Apple’s popularity (0 to 100, relative to other searches, not a count of searches) and its suggestions. Add puts a keyword in an ad group; Block makes a search a negative keyword.
Automation: the AI plan and the autopilot
Automation → Apple Ads, for the app picked at the top. Analyze now writes a plan from the last two weeks, weighed against the last eight, as the scheduled plans do; above it, when the next one comes and why. The period picker over Campaigns sets only the tables below it. A plan says: which bids to raise or lower, keywords to pause or add from the searches that worked, searches to block, budgets to change, and countries worth trying. Each step says why, and nothing changes until you apply it.
Seasons. Each plan also reads the year: the app’s new customers and trials by week from RevenueCat, Apple’s weekly popularity of its keywords in its countries, and your occasions. Before a season the app had last year (a feast, the new year), it raises budgets and the best keywords’ bids, within your goals, each step with the day it ends; Bridge takes each one back that day by itself, unless you changed it since. It adds a seasonal search as a keyword only where the data shows people searching it for an app like yours.
Goals and autopilot say what it aims at, in the account’s currency: a target cost per trial, a target cost per install, or a target return on ad spend (what the users a keyword brings have paid over what it cost; when set, bids are judged by it first, and a young keyword is given time, as subscribers keep paying); the lowest and highest bid; the largest bid and budget change; a budget ceiling for all Apple campaigns and per campaign; and how often the AI analyzes on its own.
The autopilot is off until you turn it on, for the apps and the kinds of step you tick: bids, keywords, blocked searches, budgets, pausing campaigns. After each scheduled plan it makes the plan’s steps within your goals. It never switches a campaign on, leaves a keyword’s bid and a campaign’s budget alone for 72 hours after a change, makes at most 40 keyword changes a campaign a week, and retests only keywords it paused itself, after at least 14 days. Every change, by the autopilot, from Automation or from the connector, is listed under Changes Bridge made in Apple Ads and can be undone.
New campaigns
Automation → Apple Ads → Plan new campaigns, or Plan new campaign at the top, in your own words. Give the countries and a budget a day for all of them, and the AI drafts the classic set from the App Store listing and Apple’s keyword suggestions:
- Brand: the app’s own name, exact match, bid modestly.
- Category: what the app does, in the words people search, exact match, in ad groups by theme.
- Competitor: other apps’ names, only when your notes allow it.
- Discovery: Search Match and broad match, to find new searches, blocking the other campaigns’ keywords so they do not bid against each other.
One campaign of each kind for all the countries, split where their languages differ (with keywords in each country’s
language), or a campaign per country when you tick it; at most 12 a plan. Bridge names each one like
MyApp - Category - US GB - 1234567890: the app, the kind, the countries and the App Store id, so apps sharing an
account stay apart. Look the draft over, then Make these campaigns, paused: they are made on search results with
manual cost-per-tap bids, and nothing spends until you switch one on in Apple Ads.
Good to know
- Apple gives a keyword’s popularity from 0 to 100, relative to other searches; it never gives search volume as a number.
- Search results campaigns show the app’s App Store page: there is no ad or creative to make for them.
- Apple’s spend counts in Analytics and Campaigns like Meta’s and Google’s, from the first day Bridge read, 90 days back from connecting.
- Bridge uses Apple’s Ads Platform API, which replaces the Campaign Management API v5 that Apple switches off on 26 January 2027: nothing changes for you then.