Google App Campaigns: Setup and Optimization

Last updated: October 4, 2026

1. Executive Summary

Google Ads App campaigns in 2026 are a fully automated, AI-driven format that serves ads across Google Search, Google Play, YouTube, Discover, and the Display Network from a single campaign. The system assembles creatives, selects audiences, and adjusts bids in real time—but it performs only as well as the inputs you give it. Five principles define best practice:

• Match campaign type and bid strategy to your actual business goal. Use App campaigns for installs (ACi) with target CPI when acquiring users at scale, App campaigns for engagement (ACe) with target CPA when driving post-install actions, and pre-registration campaigns to build demand before a Google Play launch. Never use an install-focused campaign when your KPI is downstream revenue.^[1][2]
• Feed the machine enough data before touching anything. Google’s guidance requires at least 100 conversions before making major edits, and in-app action events should be completed by at least 10 different users per day to give the algorithm usable signal.^[3][4] Budget must be set high enough to generate that volume without exhausting daily spend.
• Maximize asset variety across every format and dimension. The campaign assembles ads dynamically from your asset pool. A thin pool means fewer combinations, less auction eligibility, and slower learning. Upload up to 10 text assets, 20 image assets, and 20 video assets per ad group, covering multiple aspect ratios and lengths.^[4]
• Build a clean, redundant measurement stack before launch. Use Firebase or a Google-approved mobile measurement partner (MMP) such as AppsFlyer or Adjust as your primary attribution source, layer SKAdNetwork and on-device measurement for iOS, and align conversion windows across every system before the campaign goes live.^[2][5]
• Optimize to value, not volume. Raw install count is a lagging proxy for business performance. The strongest accounts optimize toward high-quality in-app events—purchases, subscriptions, or retained engagement—and use incrementality testing quarterly to validate that attributed conversions represent true lift.^[6][7]

2. Benchmarks and Numbers at a Glance

Metric	Typical range or threshold	Applies when	Source
Minimum daily budget — install or pre-registration campaigns (tCPI)	50× target CPI (e.g., $5 tCPI → $250/day minimum)	ACi tCPI and ACpre tCPI campaigns; vendor claim	[2][3]
Minimum daily budget — in-app action campaigns (ACi tCPA)	10× target CPA (e.g., $20 tCPA → $200/day minimum)	App campaigns for installs using tCPA bidding; vendor claim	[3]
Minimum daily budget — engagement campaigns (ACe tCPA)	15× target CPA (e.g., $30 tCPA → $450/day minimum)	App campaigns for engagement; vendor claim	[3]
Minimum conversion volume before major edits	100 conversions per campaign	All App campaign types during learning phase; vendor claim	[3]
Minimum daily in-app action completions for event eligibility	10 unique users per day completing the event	In-app action selection for optimization goal; vendor claim	[4]
Recommended starting tCPA vs. observed CPA	Set tCPA 20% above observed CPA (e.g., observed $25 CPA → start at $30 tCPA)	New campaigns where historical CPA data exists; vendor claim	[8]
iOS bid premium over Android	1.5× the Android tCPI or tCPA (e.g., $2.00 Android tCPI → $3.00 iOS tCPI)	Separate iOS and Android campaigns; vendor claim	[8]
Maximum budget or bid change increment during optimization	No more than 20% change at a time	Any live App campaign; vendor claim	[1][3]
Maximum asset counts per ad group	10 text, 20 image, 20 video assets	All App campaign types; vendor claim	[4]
Recommended video asset length	10–30 seconds per video asset	All App campaign types; vendor claim	[4]
tCPI premium for more-targeted campaigns vs. all-users campaigns	At least 20% higher tCPI on the targeted segment campaign	When running parallel campaigns with different audience scopes; vendor claim	[9]
tROAS adjustment to increase conversion volume	Gradual reduction of tROAS target allows access to more auctions	tROAS campaigns where volume is constrained; vendor claim — no numeric lift figure available	[10]

3. When to Use App Campaigns

App campaigns are the correct choice whenever your primary objective is to drive installs, in-app actions, or pre-launch registrations for a mobile app distributed through the Google Play Store or Apple App Store. They are not appropriate for web-only lead generation, local store visits without an app component, or brand awareness campaigns where no app action is being measured.^[1][6]

Within the App campaign family, Google offers three campaign sub-types, each suited to a distinct business stage:

• App campaigns for installs (ACi): Use when the priority is acquiring new users at scale. The algorithm optimizes for volume of installs at or below your target CPI. This is the correct starting point for apps in growth mode or new to the Google Ads ecosystem.^[2]
• App campaigns for engagement (ACe): Use when you have an existing user base and want to drive specific post-install actions—re-engagement, lapsed-payer reactivation, subscription upsells, or feature adoption. ACe requires a minimum existing user base and correct deep-link setup to function correctly.^[6][11]
• App campaigns for pre-registration: Use before a Google Play app launches. These campaigns build demand and collect pre-registrations so the app begins its public lifecycle with an existing interested audience. Apple App Store pre-orders require a separate approach outside Google Ads.^[6][12]

A critical structural rule is that you should not change the campaign sub-type after launch. If your objective shifts—for example, from installs to in-app purchases—create a new campaign rather than editing the existing one, because changing the optimization goal mid-flight disrupts learning and typically degrades performance for weeks.^[1][7]

When deciding between ACi and ACe, ask whether the users you want to reach already have the app installed. If yes, ACe is correct. If no, ACi is correct. Running both simultaneously is valid and common for mature apps, provided you exclude existing users from ACi campaigns to avoid overlap and budget waste.^[6]

4. Campaign Goals and Bidding (tCPI, tCPA, tROAS)

Bid strategy selection is the single most consequential structural decision in an App campaign. The wrong bid strategy signals the wrong objective to Google’s algorithm and produces installs, actions, or value that don’t match the business goal. As of October 2026, Google offers three primary bidding strategies for App campaigns, plus two newer goal modes that remove the requirement to set a specific target.^[2][13]

Bid strategy	Optimizes toward	Use when	Minimum daily budget
Target CPI (tCPI)	App installs at or below target cost per install	ACi or pre-registration; primary KPI is install volume	50× tCPI[2][3]
Target CPA (tCPA)	A specific in-app action at or below target cost per action	ACi or ACe; primary KPI is post-install event volume	10× tCPA (ACi) or 15× tCPA (ACe)[3]
Target ROAS (tROAS)	In-app revenue value relative to ad spend	ACe or ACi with purchase events and reliable revenue values; primary KPI is return on ad spend	Set to support adequate conversion volume at target
Optimize installs without target CPA	Installs with maximum flexibility; no specific CPA constraint	New campaigns with no historical CPA data; vendor claim — available via API as of late 2026	Set budget to drive meaningful daily volume
Optimize total value without target ROAS	Total in-app value with maximum flexibility; no ROAS constraint	Accounts with revenue events but uncertain ROAS targets; vendor claim — available via API as of late 2026	Set budget to drive meaningful daily volume

Setting the initial target CPI or CPA: Google recommends setting your initial tCPA at 20% above your observed CPA from historical data.^[8] If you have no historical data, start with a tCPI or tCPA that reflects your business’s lifetime value model—what you can afford to pay for an install or action while remaining profitable—and expect to refine it after the first 100 conversions.^[3]

iOS vs. Android targets: Google explicitly recommends setting iOS bids at 1.5 times the equivalent Android bid to account for the lower signal fidelity on iOS post-ATT and the generally higher cost of iOS users.^[8] Run iOS and Android in separate campaigns wherever possible to apply this premium without averaging down your Android spend.

tROAS best practices: tROAS is appropriate only after you have a stable volume of purchase or revenue events with consistent values flowing through your measurement stack. Google’s guidance notes that gradually reducing the tROAS target can increase conversion volume by qualifying for more auctions, which is useful when you are volume-constrained.^[10] Do not set tROAS before you have reliable revenue event data, because the algorithm will optimize toward an unreliable signal.

New no-target modes (October 2026): The Google Ads API now supports OPTIMIZE_IN_APP_CONVERSIONS_WITHOUT_TARGET_CPA and OPTIMIZE_TOTAL_VALUE_WITHOUT_TARGET_ROAS. These are useful for new campaigns where setting a target would be guesswork, but they should transition to a target-based strategy once sufficient conversion data is available.^[13]

5. Creative Assets and Asset Variety

App campaigns generate ads dynamically by combining assets from your uploaded pool. The system evaluates which combinations of text, image, video, and HTML5 assets perform best for each placement, user, and moment. A small or homogeneous asset pool produces fewer unique ad combinations, limits auction eligibility across placements, and slows learning. Maximizing asset variety is therefore not optional—it is a core performance lever.^[2][3][14]

Asset counts and format requirements: Google’s guidance allows up to 10 text assets, 20 image assets, and 20 video assets per ad group.^[4] Fill every slot. Leaving asset slots empty is equivalent to voluntarily reducing the campaign’s creative surface area.

Text assets: Write each line as a standalone phrase that makes sense in any order and communicates a single distinct benefit or call to action. Vary length—some lines short for constrained placements, some longer for expanded placements. Avoid repeating the same message across multiple text lines, because the system will favor variety when assembling combinations.^[2][3][7]

Image assets: Upload images at multiple aspect ratios—at minimum 1:1 (square), 1.91:1 (landscape), and 4:5 or 9:16 (portrait). Use high pixel density, full-frame photography or app UI screenshots, tight cropping for character or UI close-ups, and minimal overlaid text or logos. Avoid heavily filtered, blurry, or upside-down imagery, which Google flags as quality issues that reduce delivery.^[2][7]

Video assets: Upload multiple videos in the 10–30 second range^[4], covering portrait (9:16), landscape (16:9), and square (1:1) formats. Portrait video is essential for YouTube Shorts inventory. Lead with branding and a clear hook in the first 3–5 seconds. Include audio; Google’s guidance notes that audio elements can improve install rates in creative tests.^[7][15] Do not upload only one video in one format—single-format video pools foreclose entire placement categories.

HTML5 and playable assets: For eligible game campaigns, HTML5 playable assets allow users to interact with a mini-game experience before installing. Validate every HTML5 file through the Google Ads HTML5 Validator before upload. Design the playable to surface a rewarding interaction within the first 15 seconds, with branding visible from the start.^[2][3][7]

AI-assisted asset suggestions (October 2026): Google now surfaces image and text suggestions for App campaigns, drawn from your website, Google Play store listing, asset library, and stock image sources. Image suggestions are generally available; text suggestions are in beta. Use these as a starting point for asset ideation, but supplement with custom creative that reflects your app’s specific value proposition.^[11]

Iterating on asset performance: Review asset-level performance ratings in the Google Ads interface. Replace assets rated “Low” with new creative. Do not remove assets rated “Good” or “Best” to test replacements—keep strong performers in place and add new assets alongside them. Avoid replacing large numbers of assets simultaneously, as this disrupts the creative learning the system has accumulated.^[2][12][13]

6. Measurement and Attribution (Firebase, MMPs)

App campaign measurement in 2026 requires a deliberate stack architecture, not a single tool. Google’s algorithm trains on the conversion signals it receives; if those signals are delayed, duplicated, or missing, the algorithm optimizes toward a distorted picture of performance. The measurement stack must be designed before the campaign launches, because retrofitting measurement after the fact typically means losing the learning the campaign has accumulated.^[2][5]

Primary attribution options: Google recommends using either Firebase/GA4 or a Google-approved MMP as the primary attribution source for App campaigns. The three most common choices in the US market are:

Tool	Best use case	Key strength	Key limitation
Firebase / GA4	Teams wanting Google-native measurement and cross-web-app linkage	Deep Google Ads integration, free, audience building for remarketing	Limited multi-network attribution; less granular cohort reporting than MMPs[2][13]
AppsFlyer	Multi-network advertisers needing deterministic attribution, cohort analysis, and iOS SKAN reporting	Broad partner integrations, strong iOS modeling, deep-link support	Cost; requires correct SDK version and postback configuration[5][16]
Adjust	App advertisers needing cross-channel mobile measurement with partner-integrated reporting	Strong fraud protection, flexible event schema, broad partner network	Cost; iOS measurement accuracy depends on ATT consent rates[2][5]

Use GA4 as a complementary layer alongside a MMP rather than as a replacement. GA4 provides cross-web-and-app journey data and feeds audience lists back into Google Ads, while the MMP provides deterministic mobile attribution across all your paid channels.^[13]

SDK version discipline: Google’s iOS guidance explicitly requires integrating the latest version of your MMP SDK. Outdated SDK versions break event fidelity, corrupt postback handling to Google, and cause conversion undercounting that directly degrades algorithm performance.^[5][15] Build SDK version review into your monthly maintenance checklist.

Conversion window alignment: Set conversion windows in Google Ads to match the windows configured in your MMP. Misalignment—for example, a 30-day window in Google Ads and a 7-day window in AppsFlyer—creates discrepancies that make it impossible to reconcile performance data and can cause the algorithm to optimize toward events that your MMP does not credit to Google Ads.^[2][17]

Incrementality testing: Attributed conversions overstate true causal impact in any channel. The 2026 best-practice standard is quarterly incrementality testing using holdout experiments or geo-based lift studies to measure true causal lift from App campaigns. This is especially important for ACe campaigns targeting lapsed users, where organic re-engagement can be high.^[1][7]

7. In-App Events and Value Optimization

The in-app event you choose as your optimization target is the single most important signal you give the App campaign algorithm. The algorithm will literally try to find more users who complete that event. If you optimize toward a shallow event—app open, tutorial view—you will acquire users who open the app. If you optimize toward purchase or subscription, you will acquire users more likely to generate revenue. Choosing the right event requires balancing two competing constraints: the event must be meaningful enough to reflect business value, and it must be frequent enough for the algorithm to learn from it.^[2][6]

Event eligibility threshold: Google requires that the in-app action you optimize toward be completed by at least 10 different users per day.^[4] Events with fewer than 10 daily unique completions do not generate enough signal for the algorithm to learn effectively. If your primary monetization event—say, a $99 annual subscription purchase—fires only 2–3 times per day, you must temporarily optimize toward a higher-volume proxy event further up the funnel until volume grows.

Event ladder strategy: A practical framework is to define a hierarchy of events from shallow to deep and move optimization down the ladder as volume grows:

• Level 1 (highest volume): First open, tutorial start, registration — use when the app is new and deep events fire rarely
• Level 2: Tutorial completion, account creation, add-to-cart, subscription page view — intermediate proxy events
• Level 3: Free trial start, add payment method — strong intent signals
• Level 4 (highest value): Purchase, subscription activation, first revenue event — optimize here as soon as volume exceeds 10 unique completions per day^[2][4][6]

Value optimization with tROAS: Once you have a reliable revenue event flowing through your measurement stack with consistent values, switching from tCPA to tROAS shifts the algorithm’s goal from driving event volume to driving revenue value. This is the most sophisticated optimization mode and requires the highest data quality. Ensure purchase values are passed as actual USD amounts, not binary 0/1 flags, to the conversion event.^[2][13]

Event naming consistency: Event names and values must be identical across your app code, Firebase/GA4, your MMP, and Google Ads. A mismatch—for example, “purchase” in Firebase vs. “in_app_purchase” in Google Ads—creates split reporting, undercounts conversions, and degrades algorithm signal. Define a canonical event schema before any SDK integration begins.^[2][6]

Counting logic: Set purchase and revenue events to count every time they occur (not once per conversion). Set registration or install events to count once. Using “once” counting on purchase events causes the algorithm to treat repeat buyers the same as one-time buyers, which understates lifetime value and biases acquisition toward low-value users.^[2]

8. iOS, SKAdNetwork and Privacy

iOS App campaign measurement in the US operates under Apple’s App Tracking Transparency (ATT) framework, which requires explicit user consent before an advertiser can use the device’s IDFA for attribution. In practice, opt-in rates in the US vary by app category but are consistently below 50% for most gaming and utility apps, which means a significant share of iOS installs are not deterministically attributable. Google’s recommended response is a layered measurement architecture that combines consented deterministic data, SKAdNetwork (SKAN) postbacks, and on-device conversion measurement.^[5][18][19]

SKAdNetwork fundamentals: SKAN is Apple’s privacy-preserving attribution framework. When a user installs an app after clicking a Google ad on iOS, SKAN sends a postback to Google with a conversion value (configured by the advertiser) but without user-level identifiers. The postback is delayed by a randomized 24–48 hours minimum and capped at a limited number per campaign. SKAN data is therefore aggregated, delayed, and less granular than deterministic attribution—but it is the primary signal Google uses for iOS optimization when ATT consent is not granted.^[5][18]

SKAN conversion value configuration: You control how SKAN’s 6-bit conversion value (0–63) maps to your in-app events. A best-practice configuration maps the conversion value scale to meaningful revenue or engagement milestones—for example, value 0 = install only, values 1–20 = first engagement tier, values 21–40 = purchase under $25, values 41–63 = purchase $25 or above. Work with your MMP to configure and test this mapping before campaign launch, because changing it after launch resets the signal history Google has accumulated.^[5][18]

ATT prompt strategy: The timing and framing of the ATT permission prompt materially affect opt-in rates. Best practice in 2026 is to show the prompt after the user has experienced value in the app—typically after completing a tutorial, reaching a milestone, or initiating a transaction—rather than at first open. Do not show a pre-prompt screen that mimics the system dialog, as Apple’s guidelines prohibit this in the US.^[18][19]

On-device conversion measurement: Google offers on-device measurement for iOS as an additional layer that can increase the number of observed conversions without transmitting user-level data off the device. This is complementary to SKAN, not a replacement. Enable it through your Google Ads account settings if eligible.^[12][13]

iOS bid premium: Because iOS signal is noisier than Android, the algorithm requires a higher bid to operate efficiently on iOS inventory. Set iOS tCPI or tCPA at 1.5× the equivalent Android target and run separate campaigns by OS to prevent Android efficiency from masking iOS underperformance in blended reporting.^[8]

Note for non-US markets: The SKAN framework and ATT apply globally for iOS, but regional privacy regulations—including GDPR in the EU and PIPEDA in Canada—add additional consent requirements that may further reduce deterministic attribution rates in those markets. The ATT opt-in rates and measurement architecture described here are calibrated for the US context.

9. The Learning Period and Budgets

Every App campaign enters a learning period immediately after launch, after a major edit, or after a significant bid or budget change. During this period, Google’s algorithm explores the auction space, collects conversion data, and calibrates its bidding model. Performance during the learning period is typically volatile—CPA and CPI may swing well above target before stabilizing. Intervening with additional changes during learning compounds the instability and can significantly extend the time to stable performance.^[1][3][7]

What triggers a learning reset: Any of the following actions can reset or significantly extend the learning period: changing the campaign type or optimization goal, changing the target in-app event, changing conversion settings or windows, making bid changes greater than 20%, making budget changes greater than 20%, or making large structural changes to geo-targeting or asset pools during an early campaign phase.^[1][3][7]

Conversion volume threshold: Google’s guidance specifies avoiding major edits until the campaign has accumulated 100 conversions.^[3] At that point the algorithm has enough data to have formed a working bidding model, and changes are less likely to cause severe disruption. This means that budgets must be set high enough to reach 100 conversions in a reasonable timeframe—typically 2–4 weeks for healthy campaigns.

Budget sizing arithmetic: Use the minimum budget thresholds as a floor, not a target. If your tCPI is $2.00, the minimum daily budget is $100 (50 × $2.00). But if your campaign needs 100 conversions to exit learning in 14 days, you need at least 100 ÷ 14 = 7.1 installs per day, which at $2.00 CPI requires only $14.20/day in spend. In this case the 50× minimum floor ($100/day) is the binding constraint, and you should use $100/day as your launch budget. Higher budgets accelerate learning, provided the target CPI is realistic.^[2][3]

Patience and monitoring during learning: Do not pause a campaign because CPA is above target during the learning period. Pausing resets learning entirely. Instead, monitor trend direction over 7-day windows rather than daily. If CPA is declining week over week, the campaign is learning correctly. If CPA is increasing or flat after 3–4 weeks with adequate budget, investigate measurement issues before changing bids.^[1][7]

Budget change discipline: Once a campaign is live, change daily budget by no more than 20% at a time, and wait at least 7 days between changes to allow the algorithm to adapt. Doubling or halving a budget in a single change is the most common mistake that resets learning in established campaigns.^[1][3]

10. Optimization and Scaling

Once an App campaign has exited the learning period—typically signaled by stable CPA or CPI within 20% of target over a 7-day window—the focus shifts from stability to efficiency and then scale. Optimization and scaling require a disciplined sequence: first improve efficiency at current volume, then incrementally increase volume while protecting efficiency. Attempting to scale before efficiency is stable is the most common cause of CPA blow-outs in App campaigns.^[2][7]

Efficiency optimization sequence:

• Review asset-level performance ratings and replace assets rated “Low” with new creative variants. Keep “Good” and “Best” assets untouched.^[2][12][13]
• If CPA is consistently below target (e.g., actual CPA is $18 against a $22 tCPA), tighten the target by up to 20% to improve efficiency: $22.00 × 0.80 = $17.60 new tCPA. Wait 7 days before the next adjustment.^[1][3]
• If CPA is above target, diagnose before bidding. Common causes: event volume below 10 users/day, measurement gap (SDK version, window misalignment), or asset quality issues. Fix the root cause before raising the tCPA.^[2][5]

Scaling budget: Increase daily budget in increments of no more than 20% at a time with at least 7 days between changes. A campaign spending $200/day can safely move to $240/day, then to $288/day, then $346/day in successive 7-day steps—not from $200/day to $500/day in a single change.^[1][3]

Audience expansion for scale: Google App campaigns automatically target broadly, but geo-targeting restrictions can limit scale unnecessarily. If your campaign is geographically constrained to a single metro area without a business reason for that restriction, removing the geo-restriction often unlocks significant additional volume. Similarly, avoid adding audience exclusions that Google has not recommended, as they reduce the pool the algorithm can learn from.^[1][3]

Uplift Experiments for creative validation: Google now supports Uplift Experiments for App campaigns, allowing you to test new video assets against control groups at 80%, 85%, or 95% confidence intervals.^[11] Use these experiments to validate whether new creative directions actually improve install rates or in-app action rates before committing full budget to them. Run each experiment for at least 2 weeks to accumulate sufficient volume for a valid result.

Cross-channel scaling considerations: Google’s App Connect feature now enables web campaigns—Search, Shopping, YouTube, and Demand Gen—to drive clicks directly into the app.^[13] For mature apps with strong web presence, coordinating App campaign scaling with web-to-app campaign investment maximizes total acquisition efficiency. Review the web-to-app acquisition report in Google Ads to understand how much new app install volume your web campaigns are already generating before adding net-new App campaign budget.

11. Common Mistakes to Avoid

The following mistakes are the most frequently observed causes of preventable App campaign underperformance. Each stems from violating one of the core principles in this guide—and each has a specific corrective action.

• Optimizing toward an event with insufficient volume. Setting a purchase event as the optimization goal when the app averages only 3 purchases per day starves the algorithm. Threshold: 10 unique users per day.^[4] Fix: use a higher-volume proxy event until the threshold is met.
• Changing the optimization goal on a live campaign. Switching from install optimization to purchase optimization on an existing campaign resets learning entirely and wastes the historical signal accumulated. Fix: create a new campaign for the new goal and let the existing campaign wind down.^[1][7]
• Making budget or bid changes greater than 20% at a time. A single large change—doubling the budget, halving the tCPA—triggers a learning reset equivalent to launching a new campaign. Fix: change in 20% increments with 7-day gaps.^[1][3]
• Launching with a thin asset pool. A campaign with 2 text lines, 3 images (all the same aspect ratio), and 1 video cannot assemble diverse ad combinations and will underperform on placements it cannot serve. Fix: fill all available slots across multiple formats before launch.^[2][4]
• Running iOS and Android in the same campaign. Blended campaigns average bids across two fundamentally different signal environments. iOS needs a 1.5× bid premium.^[8] Blending produces under-bidding on iOS and over-bidding on Android. Fix: separate campaigns by OS.
• Misaligning conversion windows between Google Ads and the MMP. A 30-day Google Ads window against a 7-day MMP window creates a phantom conversion gap that causes the algorithm to train on data the MMP doesn’t recognize. Fix: align windows before launch.^[2][17]
• Pausing a campaign during the learning period because CPA is above target. Pausing resets learning entirely. Elevated CPA during the first 2–4 weeks is expected and is not a signal to intervene. Fix: monitor 7-day trend direction, not daily CPA.^[1][7]
• Using a single video in one aspect ratio. A landscape-only video pool excludes YouTube Shorts (requires portrait 9:16) and limits Discover inventory. Fix: upload portrait, square, and landscape videos of 10–30 seconds each.^[2][4]
• Setting event counting to “once” for purchase events. Once-counting treats a user who makes 10 purchases the same as a user who makes 1. The algorithm loses the signal needed to find high-LTV users. Fix: set purchase and revenue events to count every occurrence.^[2]
• Not running incrementality tests. Attributed conversions always overstate causal impact to some degree. Without a quarterly holdout or geo-lift test, you cannot know how much of your App campaign volume is truly incremental. Fix: schedule an incrementality test every quarter using a holdout group or geo-based design.^[1][7]

12. What Changed Recently (Last 30 Days)

The changes below reflect announcements and API release notes confirmed within approximately 30 days of October 2026. Where a feature is in beta or has limited availability, that status is noted. Treat each item as an action trigger: review your current campaign setup against the change and decide whether an update is warranted before your next optimization cycle.

New Bidding Goals Without a Hard Target

Google Ads API release notes introduced two new App campaign bidding goals that allow the system to optimize without an advertiser-specified numeric target: OPTIMIZE_IN_APP_CONVERSIONS_WITHOUT_TARGET_CPA and OPTIMIZE_TOTAL_VALUE_WITHOUT_TARGET_ROAS.^[3] Previously, campaigns optimizing toward in-app conversions or total value were required to specify a tCPA or tROAS figure before launch. These new goals remove that requirement and let Google’s bidding model set effective bids auction by auction based on predicted conversion probability or value, without being anchored to a fixed target.

The practical implication is significant for two scenarios. First, newer accounts that lack sufficient historical CPA or ROAS data to set a credible target can now launch an in-app action or value-optimization campaign without guessing a number that may be too aggressive or too conservative. Second, mature accounts running seasonal promotions — where the right effective CPA shifts week to week — can now use targetless optimization rather than manually adjusting a tCPA every few days and risking repeated learning resets.

The trade-off is reduced cost control. Without a tCPA or tROAS anchor, the system optimizes for volume or value but will not cap spend efficiency the way a hard target does. For budget-constrained US advertisers, the recommendation is to use these goals during a defined test window with a firm daily budget cap, then transition to tCPA or tROAS once the campaign has generated enough conversion history to set a data-backed target.^[2]

Expanded Web-and-App Integration

Google extended App Connect — previously limited to Performance Max, Search, and Shopping — to also work with YouTube Ads, Hotel Ads, and Demand Gen campaigns.^[13] This means clicks from those campaign types can now route users directly into the app rather than to a mobile web landing page, reducing friction in the install or re-engagement funnel. Google simultaneously released three companion features: in-product setup nudges that flag when App Connect is not configured, unified conversions that aggregate web and app conversion data in a single view, and a combined overview card in the Google Ads interface that shows web and app performance side by side.^[13]

For US advertisers running both App campaigns and YouTube or Demand Gen campaigns, this is a meaningful cross-channel measurement upgrade. The web-to-app acquisition report — also released in this window — shows how web campaigns drive new app installs and first in-app conversions, which was previously difficult to attribute cleanly without a custom MMP setup.^[13] The immediate action is to verify that App Connect is enabled across all eligible campaign types in your account and confirm that the unified conversion view is reporting correctly before drawing any cross-channel budget allocation conclusions.

AI-Assisted Asset Suggestions

Google released AI-generated image and text asset suggestions for App campaigns, drawing from the advertiser’s website, Google Play store listing, asset library, and stock image inventory.^[11] As of October 2026, image suggestions are generally available and text suggestions are in beta. The feature surfaces in the asset upload workflow inside Google Ads and is designed to reduce the time required to build an initial asset pool, particularly for advertisers launching their first App campaign or refreshing a stale creative set.

The important caveat is that AI-generated suggestions are a starting point, not a replacement for human creative judgment. Google’s own best-practice guidance emphasizes asset variety, distinct selling points, and early branding in videos — qualities that generic suggestions may not reflect for a specific app or vertical.^[7][14] The recommended workflow is to accept suggestions selectively, edit them to match brand voice and specific value propositions, and always supplement with custom assets. Do not rely solely on suggestions to fill asset slots, particularly for video, which the suggestion tool does not yet generate.

Uplift Experiments for Video Asset Testing

Google launched Uplift Experiments for App campaigns, enabling advertisers to test the incremental performance of video assets at three confidence thresholds: 80%, 85%, and 95%.^[11] Previously, App campaign asset testing was limited to the asset performance ratings (low, good, best) visible in the asset report, which reflect relative performance within the campaign but do not measure true lift against a holdout. Uplift Experiments address this by splitting traffic to measure the causal contribution of a specific video asset to install or in-app conversion volume.

The choice of confidence threshold has a direct trade-off between test speed and statistical rigor. An 80% threshold reaches a conclusion faster — requiring less traffic and a shorter test window — but accepts a 20% probability of a false positive result. A 95% threshold is more reliable but requires roughly twice the traffic to reach significance. For US App campaigns with limited daily install volume (fewer than 200 installs per day), the 80% threshold may be the only practical option that returns a result within a reasonable budget. For campaigns with higher volume, 95% is the appropriate default.^[11]

Guided Setup for App Campaigns for Engagement

Google introduced a more guided setup flow for App campaigns for engagement (ACe), with suggested inputs, updated use-case templates, and two specific audience targets highlighted in the interface: lapsed payers (users who previously made an in-app purchase but have not done so recently) and recently inactive users (users who installed the app but have not opened it within a defined window).^[11] The updated flow pre-populates bid and budget suggestions based on the advertiser’s historical in-app event data and provides guardrails around tCPA target setting that are consistent with the +20% above observed CPA rule.^[8]

ACe campaigns require users to have the app already installed, which means the eligible audience is smaller than an install campaign by definition. The guided setup now makes it easier to segment that audience by engagement history, which is a meaningful improvement for subscription apps and gaming apps with large dormant user bases. US advertisers with an existing install base of more than 100,000 users and a measurable lapsed-payer segment of at least 5,000 users should evaluate whether ACe campaigns targeting lapsed payers can generate re-engagement revenue at a lower effective CPA than acquiring equivalent new users through an install campaign.

Action Checklist: October 2026 Changes

Change	Available as of	Availability status	Recommended action	Source
Targetless bidding goals (no tCPA / no tROAS)	October 2026 API release	Generally available via API	Use for new campaigns lacking historical CPA/ROAS data; switch to tCPA or tROAS after 30 days of data collection	[3]
App Connect expansion to YouTube, Hotel Ads, Demand Gen	September–October 2026	Generally available	Enable App Connect for all eligible campaign types; verify unified conversion view is active	[13]
Web-to-app acquisition report	September–October 2026	Generally available	Review monthly to identify web campaign install and first-conversion contribution; use to inform cross-channel budget allocation	[13]
AI image suggestions for App campaigns	October 2026	Generally available	Use as a scaffold to reach maximum asset counts; edit each suggestion for brand voice and specificity before saving	[11]
AI text suggestions for App campaigns	October 2026	Beta	Test selectively; manually review all suggestions against the standalone-phrase rule before accepting	[11]
Uplift Experiments for App campaign video assets	October 2026	Generally available	Select 80% confidence threshold for campaigns under 200 installs/day; use 95% for higher-volume campaigns	[11]
Guided ACe setup with lapsed-payer and inactive-user templates	October 2026	Generally available	Evaluate for accounts with install bases over 100,000 users and a measurable lapsed-payer segment over 5,000 users	[11]

Outside the US, feature rollout timelines for the guided ACe setup and AI asset suggestions may differ; confirm availability in your target market’s Google Ads interface before building campaign plans around these features.^[11]

---

References

1. [1] https://support.google.com/google-ads/answer/9176652?hl=en support.google.com
2. [2] https://ads.google.com/intl/en_us/intl/en_us/home/grow-my-app/set-objectives/ ads.google.com
3. [3] https://business.google.com/us/accelerate/resources/articles/best-practices-for-app-campai… business.google.com
4. [4] https://business.google.com/us/accelerate/resources/articles/best-practices-guide-setting-… business.google.com
5. [5] https://support.google.com/google-ads/answer/17091671?hl=en support.google.com
6. [6] https://support.google.com/google-ads/answer/6167162?hl=en support.google.com
7. [7] https://ads.google.com/intl/en_us/intl/en_us/home/grow-my-app/evaluate-expand/ ads.google.com
8. [8] https://business.google.com/us/accelerate/resources/articles/choose-a-bid-strategy-for-you… business.google.com
9. [9] https://support.google.com/google-ads/answer/6167156?hl=en support.google.com
10. [10] https://support.google.com/google-ads/answer/6268637?hl=en support.google.com
11. [11] https://eppcdigital.com/most-complete-guide-for-google-universal-app-campaigns-includes-11… eppcdigital.com
12. [12] https://business.google.com/en-all/ad-solutions/app-ads/ business.google.com
13. [13] https://developers.google.com/google-ads/api/docs/app-campaigns/overview developers.google.com
14. [14] https://business.google.com/us/accelerate/resources/articles/make-better-creative-assets-f… business.google.com
15. [15] https://developers.google.com/google-ads/api/docs/campaigns/overview developers.google.com
16. [16] https://adapty.io/blog/google-app-campaigns-playbook-2025/ adapty.io
17. [17] https://support.google.com/google-ads/answer/9823397?hl=en support.google.com
18. [18] https://business.google.com/en-all/accelerate/resources/articles/drive-better-performance-… business.google.com
19. [19] https://business.google.com/us/accelerate/resources/articles/drive-better-performance-and-… business.google.com