On Monday, the Android campaign spent normally. On Tuesday, the cost column fell, the purchase column went blank, and the team changed its target CPA. The trouble is that those three observations do not tell you what happened first. A broken purchase event can leave the bidder with less evidence. A store restriction can stop an app from reaching users in one market. A report for a partial day can make a normal campaign look stalled. Each points to a different owner and a different next move.
If your Google App campaign is not spending, first verify that impressions and cost fell in Google Ads over comparable completed periods. Then check account, campaign, ad and app eligibility before touching bids. If the campaign is eligible, trace the specific conversion it bids toward, then test whether the budget, target or available inventory is constraining delivery. If spend is intact and only conversions are missing, start with reporting lag and measurement. Google's App campaign low-traffic guide documents these as separate reasons for limited serving.
Start with the first missing signal
Pull daily cost, impressions and the relevant conversion action from Google Ads. Keep installs or first opens as a separate row from in-app outcomes. Compare the same campaign, operating system, markets, date boundaries and account time zone before reaching for a portfolio total. A morning dashboard against a full previous day tells you little about pacing. A weighted average across Android and iOS can hide a failure isolated to one app release.
What you can see First evidence to collect First person to involve --- --- --- A new campaign has never recorded an impression Start date, account notices, ad review, selected app and eligible storefronts Campaign owner A previously active campaign drops to zero cost and impressions Last day with delivery, current statuses and timed account or release changes Campaign owner; app release owner if scope follows a store or build Cost falls while impressions continue Network mix, bid target history, budget status and receipt of the selected event Acquisition and analytics Google Ads cost stays steady but an external report shows zero Account ID, reporting period, currency, connector filters and freshness Reporting owner Spend stays steady but installs or purchases fall Conversion action, attribution date basis, store activity and event receipt Analytics or app product owner Only one network or country goes quiet Whole-campaign delivery, asset eligibility and store availability for the affected slice Campaign and creative owners
Use the table to pick where to look first. It doesn't identify a cause on its own. If impressions fell but remain nonzero, for example, a full account suspension is less consistent with the data than a restriction in part of the inventory. Old impressions in a wider date range are not evidence that the campaign is currently serving.
Put campaign changes and app changes on the same clock
In Google Ads Change History, pull changes around the first affected day: campaign and ad group status, goals, targets, budgets, geography and assets. Record when automated rules or API edits ran as well as manual changes. If Google Ads surfaces an Explanation for the affected metric and range, keep it alongside the export. Treat it as a lead; the absence of an explanation isn't a clean bill of health.
Next, ask product or analytics for the dates you cannot read from that account: app release, staged rollout, store availability, SDK or consent change, backend event deployment and provider configuration change. Timestamp each row. A purchase event that stopped reaching the partner after a release is a better lead than a bid change made two days after spend first fell. If both happened on the same day, look for an unaffected segment before assigning a cause: for example, the same event on an older app build. Use that segment to narrow the investigation, without treating it as a randomized control.
Clear eligibility before testing economics
A never-served campaign deserves a status check before anyone interprets low spend as a bid problem. Start at the account and work down to campaign, ad group and eligible ads. Check payment or account-budget notices, campaign dates, paused or missing ad groups, and ad or asset review. Confirm the campaign is associated with the intended app, and check the store listing and availability for the exact countries under investigation. Google lists account issues, review, app status and targeting among the reasons App campaigns may get limited or no traffic.
An eligible label at one level leaves the others open. The campaign can be enabled while an ad is under review; the listing can be reachable in your own country while unavailable in the intended store territory. Record the app's store ID and the tested storefront, not just its display name or a URL that resolved on your device. If you run install and engagement campaigns, record the subtype. An install campaign needs a route to a downloadable app; for engagement, the population of eligible existing users and the destination experience matter. See the App campaign goals guide if the campaign was set up for a different job than the team now expects it to do.
When an actual policy, verification or billing notice exists, route the notice to the person who can resolve it and preserve the timestamp and exact wording. Changes in budgets cannot resolve a verification request. If that is the explicit blocker, use the advertiser verification guide to assemble the right entity evidence. For assets in review, follow the review state rather than editing them repeatedly: Google's troubleshooting guide warns that edits while a review is in progress restart review. A known pending review calls for a different response from weeks of unexplained zero impressions.
Follow the exact event the campaign bids toward
When an established campaign loses delivery without an obvious eligibility change, trace the conversion the campaign uses. Google's App campaign fluctuation guide says a sharp fall in signals for the optimized action can reduce delivery and cost. That makes event receipt a delivery question as well as a reporting question. A healthy aggregate install count or an active conversion elsewhere in the account does not settle it.
Product: Can the eligible user complete the intended action on the affected build? Compare first-party counts by platform and version. An action falling in the product as well as in the media report may be a real funnel change, an instrumentation change at the source, or both.
Collection: Is the same event name and definition reaching the expected provider for this app, platform and environment? A test device that emits an event proves that the tested build can emit it; it doesn't prove that the live feed is complete.
Google Ads: Locate the exact conversion action, its app, source and selection in the affected campaign. Compare status, activity and the actual action used for bidding. Google's mobile app conversion overview distinguishes install, first open and in-app activity and documents several data sources. An action with a familiar name can still represent a different event.
Reporting: Compare the same action and date basis. Google's conversion lag reporting guide explains why recent cohorts may show incomplete conversions and misleading CPA or ROAS. If traffic and provider receipt are intact, allow for the observed lag before declaring the signal broken.
For goal setups using primary and secondary actions, verify both the setting and the goal used by the campaign. Google's conversion action documentation describes an exception: actions in a custom goal can be used for bidding even if marked secondary. Don't mark a second import primary just to see a number return. First determine whether it represents the same underlying behavior, which feed should drive bidding, and whether changing it would affect other campaigns. The Firebase versus MMP decision guide covers the longer-term ownership decision.
For iOS, annotate any relevant measurement or implementation change separately. Google's iOS troubleshooting guide describes how SKAdNetwork fluctuations and modeling adjustments can affect the read on spend and conversions. A change in one report does not prove the campaign has lost access to all its conversion signals. Ask which report moved, what the team actually changed, and what an unaffected view should still show.
Separate a budget ceiling from a target the campaign struggles to meet
A budget limits what the campaign has room to spend. A target asks the bidding system to pursue a given acquisition cost or return. A campaign consuming the available budget behaves differently from a campaign that consistently leaves headroom unused. Pull the campaign's status, relevant bid report, target history and mature outcomes. Google's App campaign best practices cover separate budget and bid constraints and point to bid strategy reports. Google also lists overly low bids among the possible reasons for limited auction participation.
For a target cost per action campaign, compare the target to recent completed acquisition cohorts for the same action, market and OS. For target return on ad spend, inspect the observed return alongside the value feed. Check for changed currency, gross versus net values, or a shift in which events carry value before reading an apparent ROAS decline as a price signal. A previous target that worked for registrations cannot validate the same target for paid subscriptions.
If the evidence points to the target, ask finance or the app owner how much the selected action can be worth, how much test spend is acceptable, and when outcomes are mature enough to judge. Set those boundaries before adjusting anything. A small increase in delivery that only clears at a cost the business cannot support is useful evidence against scaling. Increasing the budget alone cannot resolve that mismatch. If you need a different event because the original one is too infrequent, treat that as an optimization strategy decision with downstream quality checks, rather than a bid adjustment.
Check network and asset coverage without reading every missing impression as a fault
App campaigns distribute across Google inventory. A network can lose share while total campaign delivery holds. Google's network distribution guide describes how inventory availability, assets and bidding can affect distribution. Start with the campaign total, then look at where the change occurred. A YouTube decline accompanied by delivery elsewhere is different from a drop in all networks after the only eligible video and image assets were removed.
Look at the active ad group, asset review state and actual coverage by format and market. The account's asset performance labels are relative assessments within a type and campaign. A missing or pending performance rating is not the same as a policy disapproval. Check serving and review independently. If coverage is genuinely thin, use the App campaign creative assets guide to decide what format or proposition is missing. A bulk asset replacement will make it harder to tell whether the relevant issue was eligibility, network mix or creative coverage.
Write down what would prove your next action wrong
Suppose a fictional language-learning app sees a sharp spend drop after a staged Android release. The campaign is still eligible, the target and budget did not change, and first opens reach the analytics provider for both app versions. Completed lessons still appear in backend records on both versions, but the imported lesson event falls only on the new build. The campaign bids toward that lesson event.
The first action is to fix or restore the affected event path, under the app team's release and consent controls, and keep the campaign settings stable. Before the fix, record what would count as recovery: receipt of the correctly defined production event from the affected build, observable activity for the correct action in Google Ads after the expected lag, and a separate read on impressions and cost. A return of the event alongside persistently low spend would rule against treating the feed break as a sufficient explanation. If production event delivery never recovers, the test has not reached the point where a bid experiment would answer the question.
Use the same approach if the leading hypothesis is different. A corrected store availability setting should restore the intended storefront before you interpret cost. A bounded target change should have an agreed spend cap, a review date informed by conversion lag, and a criterion for stopping if mature cohort cost fails the business limit. When multiple independent problems are verified, fix hard blockers first; note which repairs happen together so you don't assign credit to one without evidence. Google notes that significant campaign changes can trigger another learning period. Avoid stacking changes while waiting for an interpretable read.
When to hand the incident to Google Ads support
If the campaign remains unexplained after you verify its scope, statuses, storefront and conversion path, prepare a short support packet. Include the campaign ID through the authorized support channel, app ID, affected territories and OS, date the symptom started with time zone, before/after impressions and cost, statuses and exact notices, last relevant changes, the action used for bidding, and the test you already ran. Say what still works. For example: “The other Android app in this account is still serving; this app is downloadable in both targeted countries; the affected campaign has recorded zero impressions in completed account days since the date below.”
Ask support to investigate the specific affected scope, rather than requesting a general account health check. Keep customer data, tokens, payment details and unrelated account exports out of the packet. Do not recreate the campaign or open a replacement account to work around a policy or account decision. Where an explicit notice provides a resolution path, follow it.
If you run paid acquisition for a live app and cannot tell whether the campaign, store or measurement stack owns the next move, talk to Sharply Labs. Bring a redacted before/after export and a short change timeline. The app growth team can review the evidence with you, identify who should own the next action, and scope one bounded paid acquisition test or a support handoff. The review does not promise approval, restored delivery, or improved performance.