Gaming and mobile apps
I connect campaigns, attribution, in-app events, retention and monetisation to understand which users carry value, which creatives work and which signals should drive optimisation.
Sector experience
On mobile, cost per install is the easiest metric to obtain and the easiest to misread. Campaigns are very good at producing cheap installs: the problem is that part of those users opens the app once and never returns, having consumed budget and taught the algorithm to find more of the same.
I work on the connection between acquisition and product. That means going into the Firebase event architecture, checking that the events sent to the platforms are reliable and frequent enough, and moving optimisation from install volume towards activation, a key event or generated value.
I have worked on gaming and mobile apps with Tiny Wizard as well, in a context where acquisition, product analysis and monetisation had to be read together. Useful decisions appear when marketing data and product data stop being two separate worlds.
Recurring problems
Situations I find in apps with paid acquisition, both in games and in subscription products.
The campaign learns from install volume and finds users who are cheap to acquire, not users who stay.
Signals that are too rare or too far from the install don’t give the algorithm enough data to learn.
Inconsistent names, missing parameters and duplication make everything built on top unreliable.
Ads, Firebase and the attribution platform show different numbers and nobody defines the source of truth.
Product data exists and gets reviewed, but never influences budget and targeting decisions.
Creatives are compared on clicks and installs, not on the behaviour of the users they bring.
Budget grows while product activation and retention can’t yet sustain the volume.
Countries with completely different value per user are pushed with the same cost goal.
User journey
Each step carries a different economic value: stopping at the first one means optimising on the least informative part of the chain.
Once a reliable and sufficiently frequent downstream event exists, optimisation should move there: continuing to optimise on installs means choosing the weakest available signal.
User journey
The creative reaches an audience: this measures pressure, not results.
Declared interest, still disconnected from the quality of the user who will arrive.
Icon, screenshots, video and reviews decide how much demand turns into an install.
The first measurable campaign-side event, and also the most overrated one.
Not every install becomes an open: the gap is already a quality signal.
The moment the user completes the action that makes the product understandable and useful.
Progression, level reached, content consumed or feature used: the signals worth optimising for.
Return at day 1, 7 and 30: this is where real users separate from noise.
In-app purchase, subscription or ad revenue: the first actual economic value.
Cumulative cohort revenue, the only level at which campaigns and markets can be compared.
Events and attribution
Value isn’t created in the ad platform: it is created in the product and has to travel back with a consistent structure.
Data path
Names, parameters and trigger conditions decided together with the product team.
Events are sent to Firebase and the attribution platform with the same semantics.
Checks on volume, duplication, delays and differences between iOS and Android.
A shared rule on windows and priority, so it is clear which source gets credit.
Choosing the optimisation event based on frequency, time distance and correlation with value.
The selected event and its value go back to Google Ads as a conversion.
Choosing the event is a strategic decision, not a technical one: an event that is too rare blocks learning, one too close to the install brings the problem back to where it started.
Areas of work
The perimeter changes between a game with in-app purchases and a subscription app, but these are the areas.
Systems and data
On mobile no single platform holds the truth: the work consists of making them talk to each other.
KPIs
Cost per install is there to monitor buying efficiency, not to establish whether the investment is working.
Two campaigns with the same cost per install can differ several times over in value per user. That difference only becomes visible after the first week.
Frequent mistakes
Method applied
The phases are the ones I use on every project, applied to the user lifecycle and to mobile measurement constraints.
Monetisation model, markets, platforms, user lifecycle and product maturity.
Validation of events, attribution rules and differences between sources.
Where it breaks: creative, store, activation, retention or monetisation.
Quick fixes on signals and spend, structural work on events and integrations.
Hands-on campaign and measurement work, coordination with product and development.
Verification on cohort value and user quality, not on install volume.
Related
Event architecture, platform integration, automation and dashboards connecting marketing and product.
Measurement, tracking and MarTech — Open the pageA different model with a similar problem: choosing the right signal instead of optimising on the first available event.
Online services and lead generation — Open the pageHow I diagnose where the journey breaks and how I set priorities.
Working method — Open the pageDirect answers
With app campaigns set on a goal consistent with product maturity: installs only in the early phase, then activation, a key event or value. Creatives and markets should be handled separately, because value per user varies widely across countries.
Google Ads and acquisitionActivation rate, day 1, 7 and 30 retention, cost per key event, payer conversion rate, average revenue per user, lifetime value and ROAS by cohort. Cost per install stays an efficiency indicator for the buy.
How I define KPIsBecause it measures what it costs to get the app installed, not what the installer is worth. Cheaper sources often bring users who never open the app or never return: the apparent cost is low, the cost per active user is much higher.
Diagnosis and prioritiesBy linking the two accounts and importing the relevant Firebase events as conversions, after validating their volume and reliability. The technical link is the easy part: the hard decision is choosing which event deserves to drive optimisation.
Tracking and integrationsBy weighing three things together: how frequent the event is, how long after the install it happens, and how strongly it correlates with future value. A perfect but rare event stops the algorithm from learning; a frequent but meaningless one brings the problem back to the install.
Event architectureBy install cohort, following the same group of users over time and observing return and cumulative revenue at defined intervals. Comparing calendar months instead of cohorts produces numbers that move with acquisition mix, not with product quality.
Analysis and dashboardsBy first deciding which source is authoritative for each decision and documenting attribution windows and rules. The three systems measure different things with different logic: fully aligning them isn’t possible, making them comparable is.
Measurement and attributionBy unifying event semantics, checking duplication and sending delays, aligning windows and time zones, and monitoring the gap over time as if it were a metric. A stable gap is manageable, a growing gap points to an implementation problem.
Data validationBy comparing the acquisition spend of a cohort with the revenue that same cohort generates at defined intervals, and comparing cohorts at the same level of maturity. A recent cohort will always show a lower return: that doesn’t mean the campaign performs worse.
Reading economic dataBy moving the optimisation signal further downstream, excluding sources with low activation or retention, and aligning the creative promise with what the user actually finds in the app. Creatives that overpromise bring cheap installs and fast uninstalls.
Traffic qualityBy comparing them not only on clicks and installs, but on the behaviour of the users they generate: activation, retention and value. An asset with a higher cost per install can be the one bringing the better users.
How I set up testsWhen the event is correctly implemented, occurs often enough for the campaign to learn, and shows an observable correlation with value. Before that point you only get unstable learning and inefficient spend.
Signals and optimisationTell me how your app works, how it monetises and which events you already have in place. The first conversation is there to check whether the signal driving your campaigns is the right one.