Google App campaigns hand distribution, bidding, and creative assembly to Google's systems. What stays with you is the part that decides whether the spend produces a business: which campaign job you are buying, which conversion action you optimize toward, whether that conversion is measured reliably, whether your assets can cover Google's inventory, and whether you can reconcile platform numbers with first-party revenue and retention.
Short answer: for Google Ads for apps campaign goals, choose the campaign subtype first, then the deepest trustworthy optimization signal actually available to your account, and judge the result with first-party value. Installs, in-app actions, and value are different acquisition optimization choices, not three interchangeable campaign subtypes. Engagement addresses existing users; pre-registration addresses an eligible Android launch. A deeper event is not automatically a better choice if its measurement, timing, volume, or account eligibility cannot support it.
The decision has five separate parts: subtype defines the job; optimization focus defines what the system seeks; bidding availability determines which controls the account can use; data readiness determines whether the signal can be trusted; and business KPI determines whether the spend deserves to continue. Google's current types page and bid-strategy table document the mechanics. The readiness and commercial evaluation framework below is editorial guidance, not a Google eligibility rule.
What Google App campaigns automate — and what stays your responsibility
Google documents three App campaign subtypes: installs (ACi), engagement (ACe), and pre-registration, which is Android only. ACi acquires users; ACe re-engages people who have installed or used the app; pre-registration builds interest before an eligible Google Play release (About different types of App campaigns). Do not describe “value” as a fourth subtype. It is an optimization focus with its own bidding and measurement requirements.
App campaigns distribute automatically across Search, YouTube, Google Play, Discover, the Google Display Network, AdMob, and other publishers that host app ads. Assets affect which ad formats can be assembled and eligible to serve. This is potential inventory, not a promise of a particular placement, network split, or acquisition cost (App campaign network distribution, About assets and ads in App campaigns). You still own the conversion definition, bid and budget decisions, market scope, creative/store handoff, and assessment of user quality.
The same automated system can report abundant installs while the business receives little value. That objection deserves an economic answer, not a claim that automation is inherently good or bad. Define what a useful new customer does after installation, how soon it happens, and which first-party records establish the value. Without those definitions, a lower CPI may simply describe a different mix of users.
If you are choosing a channel rather than a Google optimization goal, use Google App campaigns vs. Meta ads for apps or Apple Search Ads vs. Google App campaigns. If you need the whole operating system, start with mobile app user acquisition strategy. This page owns the narrower subtype, signal, and bidding-readiness decision.
Decision matrix: separate the campaign job from the optimization goal
Google's current bid-strategy table lists tCPI for ACi; tCPA for ACi and ACe; target cost per pre-registration for Android pre-registration; Maximize conversions (installs) for ACi; Maximize conversions (in-app actions) for ACi; Maximize conversion value for ACi; and tROAS for ACi and ACe. Maximize conversions (in-app actions) and Maximize conversion value are currently beta features with limited advertiser access. Their appearance in documentation does not establish availability in your account. Confirm the subtype, setup, and options actually shown before planning a migration.
The evidence and disqualifying conditions below are practical decision gates, not universal minimum conversion thresholds. Documented subtype requirements still apply.
Goal Campaign subtype / job Bid / optimization option Minimum evidence needed Business KPI Disqualifying condition --- --- --- --- --- --- Acquisition installs ACi: acquire new users tCPI or Maximize conversions (installs), subject to setup Reliable install measurement, reviewed store/onboarding path, downstream cohort baseline Cost per qualified activation and eventual contribution, with CPI as an operating metric No reliable install readback or no way to judge acquired-user quality Acquisition in-app action ACi: acquire users likely to complete a selected event tCPA; Maximize conversions (in-app actions) only where the limited-access beta is available Validated event, relevant conversion window, recurring observable signal, demonstrated relationship to value Cost per meaningful action and retained/paying cohort quality Event is duplicated, too delayed, too sparse, or economically uninformative Acquisition value ACi: acquire users with measured conversion value tROAS; Maximize conversion value only where the limited-access beta is available Trustworthy event values and currencies, mature comparable history, first-party reconciliation Realized contribution, payback and cohort revenue; reported ROAS as a defined platform metric Inconsistent values, incomparable history, immature lag, or unavailable bidding option Engagement ACe: return existing users to useful in-app activity tCPA or tROAS, subject to setup and eligibility Eligible installed-user setup, working deep links, action/value tracking, a credible incremental-return hypothesis Incremental contribution or meaningful return activity, not cheap attributed conversions alone Wrong user job, broken deep links, unmet eligibility, or inability to assess existing-user value Pre-registration Pre-registration: build interest before an eligible Android / Google Play launch Target cost per pre-registration Compliant Play pre-registration setup, release plan and post-launch measurement plan Cost per pre-registration followed by launch activation and value iOS, app already installable in the target country, or unmet Play/release requirements
For engagement, Google documents the existing-user purpose in its types guide; app engagement ads rely on deep links to the appropriate in-app destination. Test those links rather than assuming a generic app open satisfies the promised action. Current engagement eligibility must be checked in the account; this table does not replace it.
Pre-registration is not an inexpensive substitute for installs. Google's dedicated pre-registration guidance requires an APK uploaded to at least one Play Console release track, launch within 90 days after making the app available for pre-registration, and an app that is not available for install or purchase in the targeted country. It is Android/Google Play only. Judge the later transition from registered interest to real usage; registrations alone do not demonstrate paying demand.
The goal-readiness ladder: installs, in-app action, value
Google offers bid strategies aligned to installs, in-app actions, and value, depending on subtype, setup, and account availability (Choose a bid strategy for your App campaign). Choosing among them is a readiness question, not a preference question. Ask four things about the event you want to optimize toward.
Is it meaningful? The event has to be a step people who become valuable actually take, and one that people who churn largely do not. Do not assume that “tutorial completed,” “trial started,” “first order placed,” or “profile connected” passes this test. Compare each candidate with later retention and realized value in your own product.
Is it measured reliably? The event must fire consistently, be attributed to the campaign through your tracking setup, and survive app updates. Google documents app conversion tracking setup and supported measurement routes; use the route appropriate to your app and validate the imported conversion action (Set up mobile app conversion tracking). If the event is only sometimes recorded, optimization is learning from noise.
Is it timely enough? Conversions that land far outside the campaign's conversion window are not usable as an optimization signal in the same way as near-term ones. Google defines conversion windows for App campaigns explicitly, and how you set them changes both optimization and reporting (About conversion windows for App campaigns). A 30-day-to-purchase product optimized against a 7-day window is optimizing against a truncated view of itself.
Is it aligned with economics? The event should correlate with money. If your "activated" users monetize no better than the rest, deepening the optimization goal will not fix the economics; it will just change which cohort you overpay for.
Fail any one of the four and you step down a rung. Fail reliability and you should be fixing instrumentation before touching bids. Fail timeliness and either shorten the event you optimize toward or accept installs as the working proxy while you build a better one. This framework deliberately does not prescribe a hard conversion-volume threshold. Verify Google's current requirements for the selected setup separately; a generic number is not evidence that your event is reliable or commercially meaningful. Observe whether signal arrives consistently, whether mature cohorts can be compared, and whether delivery remains interpretable at your spend level.
The same logic applies on Meta with different mechanics; if you run both, our piece on Meta app campaigns and post-install events covers that side.
A staged migration path: installs → in-app action → value
Stage one: establish an install baseline without mistaking it for success. Confirm that installs are recorded through the chosen measurement route, the app/store handoff works, and cohorts can be followed into activation, retention, and realized revenue. Write down the conversion window, market, platform, event definitions, and evaluation period before changing optimization. A baseline is a comparison reference, not permission to scale low-quality volume. If even install reporting cannot be trusted, repair instrumentation before relying on automated conversion optimization.
Stage two: move to a meaningful in-app action when the signal survives scrutiny. Select an event with an observable relationship to value, consistent firing, tolerable delay, and enough recurring activity to evaluate. Check whether tCPA is available for the intended ACi setup; do not design the launch around Maximize conversions (in-app actions) unless that limited-access beta is actually enabled (bid-strategy guidance). Keep market scope and evaluation definitions stable enough to distinguish a goal change from a simultaneous audience, pricing, or onboarding change. Record the change date and judge mature cohorts rather than the first incomplete reporting day.
Stage three: move to value only when the value itself is trustworthy. Validate the selected event's value, currency, timing, duplicate handling, and relationship to first-party receipts. A trial, gross purchase, and net contribution are different quantities. Name the quantity you send and the quantity the business will use for its decision. Confirm tROAS availability or, where applicable, limited-access Maximize conversion value; neither removes the need for reliable values (bid-strategy guidance).
For an initial tROAS, Google directs advertisers to identify the biddable event's conversion window, examine a comparable campaign with the same App ID, biddable event, location, and language, and use achieved conversion value/cost over that window as a recommended starting reference. Account for conversion lag rather than treating recent incomplete returns as mature performance (Set an initial Target ROAS). Keep conversion/lookback-window definitions and the evaluation period comparable too. A different app, event, geography, language, or window is not interchangeable evidence. This is a starting-reference method, not a guaranteed target or permission to import another app's ROAS benchmark.
Rollback rule: diagnose the signal before “fixing” the target. If event counts disappear, values jump without a business explanation, or deeper-event activity becomes too sparse to interpret, stop increasing spend into uncertainty. Inspect app releases, event imports, duplication, consent-dependent measurement, conversion definitions, reporting lag, and first-party records. Google's fluctuation troubleshooting guide supports configuration and measurement checks; the commercial rollback rule is our editorial recommendation. If the deeper signal remains unusable, consider the last trustworthy shallower acquisition goal or pause the affected spend while measurement is repaired. Do not switch blindly to installs if installs are also broken, and do not use a rollback to disguise worsening user economics.
What Audience signals do — and what they do not restore
Google's current types page describes Audience signals as an optional input for App campaigns for installs: information supplied to Google AI about the users most likely to install the app. Explain it as input to automation, not a manually enforced audience boundary, guaranteed targeting, or a promise of better customers. It does not turn ACi into a manually targeted campaign.
A useful planning question is whether your proposed signal describes people with credible product relevance. A separate measurement question is whether the resulting cohorts become valuable. Those questions must stay separate: supplying a signal does not establish causal lift, exclude every other user, or solve poor conversion tracking. Engagement's existing-user job is also not recreated by attaching Audience signals to an install campaign. Select the correct subtype before debating the optional input.
An implementation workflow that survives contact with reality
Define the business event before the ad account. Write down, in one sentence, the in-app action that predicts revenue for your app, and how long after install it typically happens. This single decision drives the campaign job, the bid strategy, and the conversion window.
Configure and validate conversion tracking. Set up app conversion tracking through your chosen path and confirm the conversion is importing and counting as expected before spending meaningfully (Set up mobile app conversion tracking). Validate on both platforms and both a fresh install and an update path.
Choose the campaign job. Installs for acquisition, engagement for re-activating an existing base, pre-registration only for an unreleased Android title (About different types of App campaigns, pre-registration requirements).
Set the optimization goal one rung below your ambition. If you want value optimization but revenue import is untested, start at the in-app action and move up when the data holds.
Assemble varied, eligible assets. Cover text, image, video, and where relevant HTML5, in the formats and orientations Google accepts, so the campaign can serve across networks (About assets and ads in App campaigns). Missing formats can narrow eligible inventory; check your actual asset coverage rather than assuming a universal constraint.
Launch with controlled scope. Constrain geo and language to markets whose store listing, pricing, and onboarding you have actually reviewed. Document changes so reporting remains interpretable; Google's troubleshooting guidance identifies configuration changes and learning as possible contributors to fluctuations.
Inspect signal health before judging performance. Are conversions importing daily? Is the conversion window consistent with reality? Are assets eligible and serving?
Reconcile. Compare Google Ads conversions against your analytics or MMP and against first-party revenue and retention. This is a separate discipline, described below.
Asset coverage: what "let Google assemble it" really requires
Because ads are assembled from assets, coverage affects eligible formats; it does not determine a guaranteed reach or network allocation. Text lines carry Search and Play surfaces. Landscape, portrait, and square images plus video in multiple orientations open YouTube, Display, and Discover inventory. Engagement campaigns have their own asset requirements distinct from install campaigns (Asset requirements for App campaigns for engagement). For the portfolio-level view — which concepts and formats to produce and how to read the asset reporting — see Google App campaign creative assets: build a portfolio, not a file dump.
Two disciplines matter here. First, variety is not volume: five near-identical videos give the system less to work with than three genuinely different concepts. Second, asset-level reporting tells you which assets are contributing, not why — so use it to investigate relative underperformance and to inform your next production round rather than as a substitute for structured creative testing. Our creative testing framework covers that separately, and store-side conversion is its own lever, covered in app store screenshots and paid UA.
One caution: nothing about adding assets guarantees distribution on a given surface. Adding eligible formats makes serving possible; the auction still decides.
Why did delivery or cost change? A diagnostic order of operations
Google publishes guidance on troubleshooting App campaign fluctuations, and the useful habit is to rule out measurement and configuration before rewriting strategy (Troubleshoot App campaign fluctuations).
Conversion pings and reporting delay. Are conversions still arriving, and is today's number simply incomplete? Do not judge mature economics from a partial day.
Conversion window definition. A changed or misaligned window shifts both what the system learns from and what you see (About conversion windows for App campaigns).
Bid and budget constraints. A target that the auction cannot meet suppresses delivery; a budget raised sharply changes the mix of inventory bought.
Asset eligibility and variety. Disapproved or unavailable assets may narrow eligible formats; inspect status before attributing a change to the goal.
Geo and language scope. New markets change both cost and user quality, often at once.
Store-page handoff. Click-to-install conversion is a store-side variable that ad-side reporting will attribute to the campaign.
Downstream economics. Stable CPI with falling activation is a reason to inspect cohort mix, onboarding, measurement, and goal alignment, not proof of one cause.
The measurement reconciliation triangle
Three systems will disagree, and that is expected rather than broken.
Google Ads conversions are defined by your conversion action and its window; they answer "what did Google's system see and optimize toward?" Analytics or your MMP applies its own attribution logic across channels and answers "which channel does our arbiter credit?" First-party retention and revenue answers "what did the business actually receive?" Only the third is money.
Reconcile directionally, not to the decimal. Ask whether the three move together over a cohort period, whether Google-reported conversions per install match your own event counts within a stable ratio, and whether that ratio changes when you change something. A widening gap is a signal to inspect instrumentation, not to average the numbers.
Also keep attribution and incrementality apart. Attribution assigns credit for observed conversions; it cannot tell you what would have happened without the spend. Incrementality asks that counterfactual question and needs a designed test. Our piece on mobile app attribution architecture goes deeper on how to build the system rather than argue with it.
A hypothetical calculation (illustrative only)
The following numbers are invented round figures used to demonstrate a method. They are not Sharply Labs results, client data, or benchmarks, and they are not a forecast for your app.
Two campaigns, each spending $20,000 in a month.
Campaign A Campaign B --- --- --- Spend $20,000 $20,000 CPI $2.00 $4.00 Installs 10,000 5,000 Paid activation rate 1.0% 3.0% Paid activations 100 150 Cost per paid activation $200 $133
Campaign A wins on CPI and install count and loses on the only one attached to revenue. If average first-year gross profit per paid activation is $250, A returns $25,000 against $20,000 spent and B returns $37,500 — and if you must recover cost within 90 days, the comparison depends further on how much of that $250 arrives inside the window. The arithmetic is trivial; the discipline is refusing to rank campaigns before the value column exists.
When Google App campaigns are not the right first move
The event stack is unreliable. If conversions import inconsistently, you cannot optimize past installs and you will not be able to reconcile results. Fix instrumentation first.
You cannot produce varied assets. Without video and a range of image formats, you are asking an inventory-hungry system to work with a fraction of its inventory.
You need control over narrow search intent. When a specific, high-intent query set is the whole opportunity, automated cross-network distribution is the wrong shape of tool for that job.
Onboarding or the store page is weak. Paid traffic will expose the leak faster than it fills the funnel.
You cannot reconcile value. If nobody can connect a cohort to revenue, you will be managing a dashboard rather than a business.
Limitations and tradeoffs
There is no universal best campaign type, bid strategy, or optimization goal. Automation reduces operational load and removes levers at the same time. Deeper optimization goals align spend with value but need better measurement to work at all, and they can restrict delivery when signal is thin. Shallower goals learn faster and buy volume that may not convert. Conversion windows that capture more of the truth also slow down the feedback loop. Every one of these is a trade, and the right position depends on your product's time-to-value, your instrumentation maturity, and your tolerance for slower learning. Google publishes the mechanics (current bid strategies); the judgment is yours, and nobody can promise a CPI, CPA, ROAS, or ranking outcome in advance.
Common questions
Should a new app start with installs or an in-app action?
Start with the deepest event you can measure reliably today, which may be installs for a newly launched app. A pre-launch Android app instead needs to meet the separate pre-registration requirements; an unreleased app should not be treated as an install campaign by default. Move to an in-app action when that action is instrumented, validated, and occurring close enough to install to be learnable inside your conversion window.
Can one App campaign serve both acquisition and re-engagement?
They are distinct campaign jobs and subtypes. Acquiring new users and bringing existing users back use different campaign types with different targeting and asset requirements (About different types of App campaigns, Asset requirements for App campaigns for engagement). Deciding which of the two jobs should receive the next budget is a separate allocation question, covered in App install vs. re-engagement campaigns.
Is pre-registration available on iOS?
App campaigns for pre-registration are Android and Google Play only. Check the release-track, launch-timing, and target-country conditions in the dedicated guidance. Plan iOS launch demand through other means.
Why do Google Ads and my MMP disagree?
Because they use different attribution definitions and windows. Expect a stable gap and investigate when the gap changes, rather than expecting the two datasets to match.
How much creative do we need?
Enough variety across text, image, and video formats for the campaign to be eligible across Google's networks (About assets and ads in App campaigns). Distinct concepts contribute more than duplicates of one idea.
Where this fits
Choosing the campaign job is one decision inside a larger acquisition system that includes store conversion, activation, retention, and measurement. If you are also deciding who runs it, our guide to evaluating a mobile app marketing agency sets out the diligence questions, and our mobile app growth work describes the app-specific scope. Our growth services cover the wider paid-growth engagement.
Talk it through
If you are running or preparing Google App campaigns, we will examine: the campaign job you are actually buying, the conversion action and window you optimize toward, bidding readiness and actual account availability, your asset coverage across Google's networks, the store-page handoff, and whether your first-party economics can be reconciled with platform reporting. You get a prioritized written diagnosis of what to fix first and what to leave alone.
No CPI, CPA, ROAS, ranking, or growth outcome is guaranteed, and we will say plainly if we think the channel is the wrong first move for your situation.
Sources
All sources verified and accessed 2026-10-14. Google documents product mechanics and eligibility; the readiness ladder, migration gates, business KPIs, and rollback judgments are editorial recommendations. Account availability can vary, and documentation is not a performance promise.
Google Ads Help — About different types of App campaigns. Three subtypes and optional ACi Audience signals; not a hard targeting guarantee.
Google Ads Help — Choose a bid strategy for your App campaign. Subtype-specific bid table; limited-access beta for Maximize conversions (in-app actions) and Maximize conversion value. Actual account availability must be checked.
Google Ads Help — About App campaign network distribution. Automated network distribution and asset-format coverage; not guaranteed placement or performance.
Google Ads Help — About App campaigns for pre-registration. Android/Google Play preregistration and release-track, 90-day launch, and target-country requirements.
Google Ads Help — Set an initial Target ROAS. Comparable App ID, event, location/language and conversion-window method with conversion-lag caveats; not a performance guarantee.
Google Ads Help — Set up mobile app conversion tracking. App conversion setup and measurement routes; does not establish that a specific account is correctly instrumented.
Google Ads Help — About conversion windows for App campaigns. Click/view conversion windows and counting periods; not proof of causal incrementality.
Google Ads Help — About app engagement ads. Deep-linked in-app destinations and existing-user ads; not a complete current ACe eligibility checklist.
Google Ads Help — Asset requirements for App campaigns for engagement. Engagement asset requirements; no guaranteed delivery or uplift.
Google Ads Help — About assets and ads in App campaigns. Automated assembly of supplied assets and inventory formats; asset observations are not causal creative tests.
Google Ads Help — Troubleshoot performance fluctuations and changes in App campaigns. Measurement/configuration and performance fluctuation diagnostics; does not identify a specific campaign root cause.