A game marketing campaign worked when it produced the outcome it was designed to produce at an acceptable cost. That usually means qualified pre-launch demand, newly acquired players who keep playing, purchases, or revenue from an acquired cohort—not simply a large impression count or a low cost per install (CPI).
Choose one primary outcome KPI, report its cost, and then check downstream player quality. Also separate attributed results (credit assigned by a platform’s rules) from incremental results (outcomes that would not have happened without the advertising). A campaign can look efficient in an attribution dashboard while creating little additional demand.
Which KPI should be primary for your campaign?
The campaign objective and game business model determine the primary KPI. Google’s media-effectiveness guidance says to “Choose a KPI that best represents your campaign objective.” A pre-launch PC campaign, a free-to-play mobile launch and a re-engagement push should therefore use different scorecards.
| Campaign and business model | Primary decision KPI | Supporting measures and interpretation |
|---|---|---|
| Pre-launch paid PC or Steam game | Incremental qualified interest, if measurable; otherwise campaign-linked wishlist additions and cost per wishlist | Tracked visits, eligible-visit-to-wishlist conversion, additions versus deletions, territory, and post-release purchases from wishlist cohorts |
| Post-launch paid PC or Steam game | Incremental purchases or net contribution, if measurable; otherwise attributed purchases, cost per purchase and revenue relative to spend | Steam tracked purchases and later studio financial results. Store traffic alone is not an outcome; Steam UTM reporting excludes in-game purchases |
| Mobile free-to-play launch | Players reaching a meaningful early action at an acceptable cost | Installs and CPI for acquisition efficiency; D1, D7 and D30 retention, tutorial or level progress, payer conversion, ad monetization and cohort LTV for quality |
| Mobile monetization or re-engagement | Incremental or attributable cohort revenue relative to campaign cost over a stated payback horizon | Day-specific ROAS and LTV, purchase and ad-revenue components, retention and cohort size; segment by country, creative and platform only when data is comparable |
| Awareness campaign | Lift in the intended awareness or consideration outcome, when measured | Reach, impressions and video completion describe delivery or attention. They do not automatically prove sales |
The supporting measures explain why the primary KPI moved. Keep them in the same report, but do not let a diagnostic metric replace the business outcome.
How do attribution and incrementality differ?
Attribution allocates credit to an ad, link or campaign under a platform’s reporting rules. Incrementality estimates the additional outcomes caused by advertising by comparing an exposed group with a suitable control or holdout.
For example, a platform may credit a purchase to the last ad click even if the player was already planning to buy. A lift design asks how many purchases occurred beyond the number expected without the campaign. Google’s guide distinguishes ordinary ROAS and attribution reporting from incremental conversions and incremental ROAS in conversion-lift work; see the media-effectiveness guide.
- Use attributed results for daily optimization, creative comparisons and operational pacing.
- Use incrementality studies, geo holdouts or other controlled comparisons when the decision is whether advertising created new demand.
- Label both views in the dashboard. Never present an attributed purchase as proof that the ad caused it.
What do installs and CPI actually prove?
CPI is campaign spend divided by installs credited under a stated reporting definition. It measures acquisition efficiency, not profitability or player quality. A low CPI can be attractive while the acquired cohort churns before reaching the game’s monetization events.
Define the install event before comparing sources. Gamesight documents a default game “Install” goal based on a first-play launch event because direct download or purchase measurement is unavailable for many distribution platforms. Inspect your own event definition: a reported download, an install, a first launch and a completed tutorial are different events.
Use a qualified early action
For a mobile launch, pair CPI with an action that predicts later value for your game, such as completing the tutorial, reaching a level, starting a match or returning the next day. Report cost per qualified action as a separate ratio. State the denominator and event window so a campaign cannot improve its apparent efficiency by counting a weaker event.
How do retention and cohort LTV prove player quality?
Retention is the share of an install cohort that returns on a specified day. D1, D7 and D30 are different observations, not interchangeable labels. Document the platform’s cohort and activity definition before comparing campaigns.
Cohort lifetime value (LTV) is revenue attributed to that cohort over a stated window. Depending on the integration, it can include in-app purchases, advertising revenue and subscriptions. Google’s AdMob cohort reporting can expose total, purchase, advertising and subscription LTV alongside retention and a Google Ads app-campaign dimension; confirm which integrations and dimensions your game actually has.
Read the measures together
- Retention without revenue: players return, but the game may not monetize them or the paywall may be delayed.
- Revenue without retention: a small number of payers can inflate early LTV while the broader cohort disappears.
- Low CPI and weak D7: acquisition is cheap but likely producing low-quality traffic.
- Strong D7 and rising D30 LTV: a slower, more valuable cohort may justify a higher acquisition cost.
Compare cohorts on the same acquisition date range, territory, platform, creative and observation day. A D7 LTV cannot fairly be compared with a mature D30 LTV.
How should you calculate ROAS?
ROAS at day n is observed cohort revenue through day n divided by the spend used to acquire that cohort:
ROAS D7 = cohort revenue observed through day 7 ÷ acquisition spend.
Always name the window. D7 and D30 describe different amounts of time for the same cohort to earn revenue. Unity notes that an observed ROAS below 100% can rise as additional revenue accumulates; that does not guarantee eventual payback.
- Use matching cohorts for cost and revenue.
- State territory, platform, currency and attribution source.
- Identify whether revenue is gross, net of store fees, estimated, refunded or includes advertising revenue.
- Keep ROAS separate from net profit. Fees, refunds, customer support, development and operating costs require a contribution or profit calculation.
A campaign may be acceptable at D7 if the business can fund the time until its target payback day. Set that horizon from margin, monetization curve, budget and cash-flow constraints rather than a universal percentage.
Are Steam wishlists proof of sales?
Wishlists are evidence of interest, not guaranteed purchases. For a pre-launch Steam campaign, track wishlist additions and cost per wishlist, then follow the cohort through release and later purchases.
Valve states: “Because everyone wishlists for different reasons, there’s no formula to accurately predict sales from wishlists.” Use your own game’s historical cohorts, if available, instead of promising a fixed wishlist-to-sales multiplier. The Steam wishlist documentation also describes additions, purchases, deletions and regional views.
Understand Steam UTM boundaries
Steam distinguishes Total Visits, Trusted Visits and Tracked Visits. Only eligible tracked visits contribute to UTM wishlist, purchase or activation conversions. The UTM conversion window is 72 hours; dashboard visits update hourly, while conversions finalize after four days. Some in-app social-browser journeys cannot be connected to a later Steam action.
Steam UTM reporting does not include in-game purchases, and demo-only downloads do not count as conversions on a base-game UTM link. Its store-page traffic and store-page purchase conversion rate are not themselves factors in Steam store visibility, so do not market either metric as an algorithm hack. See the Steam UTM Analytics documentation for the reporting definitions.
Which mobile KPIs matter after launch?
For a free-to-play launch, judge whether acquired players reach the intended early action, remain active and generate value at a cost the game can recover. Installs and CPI are the starting point; retention, progression and monetization determine whether the cohort is useful.
Launch acquisition
- Installs and CPI show how efficiently media generated the defined install event.
- Early-action rate shows the share reaching the tutorial completion, first match, level or other meaningful event.
- D1, D7 and D30 retention reveal whether the traffic returns.
- Payer conversion, purchase revenue and ad revenue show monetization paths.
- Cohort LTV and day-specific ROAS connect quality to the acquisition cost.
Monetization and re-engagement
Use incremental or attributed cohort revenue relative to spend, with a stated payback horizon. Break revenue into purchases, advertising and subscriptions where the measurement system supports it. Segment by country, creative and platform only when each segment has enough comparable data; otherwise apparent differences may be reporting noise.
Apple campaign reporting
Apple campaign links can associate campaign tokens with impressions, product-page views, downloads, usage, sales and subscriptions. Individual campaign metrics appear only after Apple’s reporting threshold is met, and some usage reporting depends on user data-sharing consent. A blank or thin source breakdown therefore does not necessarily mean zero activity.
How do you build a useful campaign scorecard?
Build the scorecard before launch so the event definitions and observation windows cannot change after results arrive.
- Write the business objective. Specify whether the campaign seeks qualified pre-launch demand, first-play acquisition, purchases, revenue or awareness lift.
- Name the primary KPI and acceptable cost. For example, cost per qualified early action or D30 ROAS, with the margin or payback condition that makes it viable.
- Define the conversion event. Record the exact event, denominator, attribution window and whether it means a download, first launch, tutorial completion or purchase.
- Connect first-party reporting. Use Steamworks UTM Analytics for eligible Steam traffic and App Store Connect Analytics for Apple campaign links. Confirm integrations before promising cohort dimensions from an ad or analytics platform.
- Capture quality outcomes. Add retention, progression, payer conversion, purchase revenue, ad revenue and subscriptions when instrumented.
- Set observation days. Report D1, D7, D30 or another stated horizon consistently for each cohort. Do not compare an immature cohort with a mature one.
- Separate attribution from lift. Label platform-credited outcomes and, where the decision requires causality, run a holdout, geo comparison or other lift design.
- Cut only comparable segments. Use territory, platform, creative and campaign cuts when cohort size, currency, event definition and time window match.
Recommended dashboard fields
- Objective and target
- Spend and campaign-tagged traffic
- Conversion event and denominator
- CPI or cost per qualified action
- D1, D7 and D30 retention where relevant
- Cohort purchase, advertising and subscription revenue
- ROAS at each stated observation day
- Attribution source, eligibility rules and measurement caveats
This is a practical synthesis, not a universal official template. Keep the definitions visible beside every number so stakeholders know what the score means.
What platform and geography issues can distort comparisons?
Platform reports are not interchangeable market benchmarks. Segment actual results by territory and platform, and preserve each platform’s eligibility and consent rules.
Steam
Compare ad-platform clicks with Steam tracked visits only after accounting for the 72-hour UTM conversion window, hourly visit updates and four-day conversion finalization. A missing connection from a social-app browser can make Steam conversions look lower than click totals without proving that the campaign failed.
Apple
Campaign-level usage and sales may be sparse because of reporting thresholds and user data-sharing consent. Treat “not shown” as a reporting limitation until other first-party records establish the outcome.
AdMob and mobile cohorts
AdMob can expose different LTV components and an app-campaign dimension only when the required integrations and data are available. Confirm the dimensions for the game being evaluated before comparing sources.
Currency and proceeds
Apple analytics distinguishes reported sales from final proceeds. Use the same revenue basis across campaigns, and do not compare gross store revenue in one market with net proceeds in another.
What does not count as proof that a campaign worked?
- Impressions alone: they show delivery, not demand or revenue.
- Video completion alone: it indicates attention, not player quality.
- Cheap CPI alone: it ignores retention, progression and monetization.
- Store traffic alone: traffic can rise without purchases or incremental demand.
- Attributed ROAS alone: platform credit is not causal evidence and is not net profit.
- A wishlist multiplier: no universal wishlist-to-sales formula exists.
- A single cross-game benchmark: genre, territory, platform, price, margin and monetization curves differ too widely for a universal CPI ceiling or acceptable ROAS number.
Targets should come from the game’s economics: contribution margin, expected LTV curve, budget, cash-flow tolerance and acceptable payback period. If those assumptions change, update the target and document the change rather than moving the KPI after the fact.
How should you state the final verdict?
Report one sentence that combines outcome, cost, quality and evidence strength: what happened, how much it cost, whether the acquired cohort retained and monetized, and whether the result is attributed or incremental. A campaign “worked” only when that result meets the game’s predefined economic threshold; a platform dashboard cannot supply that threshold for you.
Frequently Asked Questions
How long should I wait before judging a game campaign?
Wait until the campaign’s chosen observation window has matured. An install-day read can assess delivery, but retention and LTV-based decisions should use the stated D7, D30 or payback horizon rather than an early snapshot.
Can I compare ROAS between iOS and Android campaigns?
Only after matching the observation day, revenue basis, territory, currency, attribution rules and cohort definition. Consent and reporting differences can make a raw platform-to-platform comparison misleading even when the game and creative are identical.
What should I do when a platform shows too few conversions to report a campaign?
Keep the platform limitation in the scorecard, aggregate only comparable periods or cohorts, and use first-party event and revenue records where available. Do not convert a blank report into a zero-result assumption.
Do Steam wishlists predict launch-day sales?
They indicate interest, but Valve says there is no accurate universal formula for turning wishlists into sales. Build a game-specific model from your own post-release cohorts instead of publishing a fixed multiplier.