A catalogue must earn its credit
A game catalogue is a decision system: which games appear, where, to whom, and under what conditions. Game count, impressions, and gross revenue can rise while players are directed to weaker choices, relevant titles are hidden, or demand concentrates among few providers.
A defensible framework follows eligible exposure through meaningful launch, post-launch play, return, and net value. Separate catalogue effects from traffic, promotions, availability, and short-term game outcomes. Treat patterns as associations until the design supports causality: more launches after a ranking change are not proof if that day's traffic mix also changed.
Exposure is allocation, not demand
Impression Share measures storefront attention:
Impression Share = Item impressions / All eligible catalogue impressions
Eligible excludes titles unavailable for a player's geography, device, currency, or regulatory state. Count an impression only when a tile enters the viewport for a defined minimum period; below-fold rendering is allocated inventory, not observed attention.
High share means opportunity, not demand: a hero slot can have high share and weak launches, while a low-share game can appeal strongly through search, filters, or a provider page. Use it to audit distribution across providers, categories, and game ages—not as demand or ranking success.
Launch rate tests the storefront promise
Launch Rate asks whether exposure produced an intentional start, at user level and within a fixed attribution window:
Launch Rate = Users who launch after a valid impression / Users with a valid impression
Exclude duplicate clicks, failed launches, and launches outside the window. Match the denominator to the reviewed surface; do not compare a homepage carousel with search unless eligibility, rank, and player intent are comparable.
Unlike raw launches, Launch Rate accounts for distribution, but repeating a popular game across placements can game it. Track unique exposed users, unique launched titles per user, and repeat-exposure frequency alongside it.
For example, 10,000 valid impressions and 900 launches yield 9%. After a placement change, 1,000 launches from 15,000 impressions yield 6.7%. Starts increased through distribution, but the tile became less persuasive per opportunity. Both matter.
Depth must reflect meaningful play
A launch may be accidental, fail during loading, or end after one round. Session depth should represent a real game experience, not time with a loading spinner.
For a casino catalogue, use:
- Successful game load rate — successful game starts divided by launch attempts.
- Meaningful play rate — launched sessions reaching a pre-defined threshold of settled rounds or staking activity.
- Rounds per meaningful session — settled rounds divided by sessions passing that threshold.
- Return-to-catalogue rate — players returning to browse after a short or failed session.
- Post-launch NGR contribution — net gaming revenue from the launch cohort after direct costs, with a stated accounting window.
Session duration can reflect engagement, an inactive tab, slow exit, or an open game state. More rounds can signal interest or different mechanics. Compare depth within game type, device, market, and lifecycle stage before judging a ranker.
The strongest signal is not maximum depth on one title, but improved meaningful launch and return without worse load reliability, player choice, or net value.
Provider concentration exposes hidden risk
A broad catalogue can have narrow effective demand. Measure concentration in displayed impressions, launches, and value. Provider share at each stage shows whether attention is broad but demand converts to a few providers, or a few control the entire journey.
Top 5 Provider Share = Launches from five largest providers / All catalogue launches
For a fuller view:
HHI = sum of squared provider shares
Use the same unit within a comparison: provider launch, Bets Sum, or NGR share. Higher HHI after a ranking release is not automatically bad; it may reflect relevance for a defined audience. Risk arises when concentration rises while discovery falls, providers lose viable exposure routes, or the catalogue depends on few titles and commercial relationships.
Segment results: new players may need recognisable titles, while established players have different preferences and search patterns. Aggregate shares can hide lost choice for one cohort.
Freshness follows a decay curve
New games attract attention because they are new, heavily placed, or meet an existing need. Track age from when a game became eligible on a surface, not global release. Calculate age-bucketed Impression Share, Launch Rate, meaningful play, and post-launch return to show week one, following weeks, and later life rather than treating titles as equally mature.
A post-launch decline is expected for novelty-slot titles. Ask whether performance stabilises above comparable games after premium placement ends. Compare within genre, provider, position band, country, device, and entry channel; a fresh live game and older slot do not share demand conditions.
Decay prevents crediting a merchandising model for gains from recently released titles. Hold game-age mix constant or report its shift.
RTP noise can impersonate retention
RTP = Player wins / Bets Sum
Observed RTP is volatile, especially in short windows or small cohorts. Favourable outcomes can raise near-term return and unlucky outcomes can shorten it; neither proves placement caused retention or that a game has a stable retention advantage.
Promoting titles after brief high returns or engagement can create a noise feedback loop: winners gain exposure, bets, and measured revenue. High-volume players further distort averages through their shares of bets and outcomes.
Evaluate retention from a fixed cohort start and clearly define the event, such as a return session with meaningful play within seven or 28 days. Examine results with and without outcome-sensitive groups, use enough settled activity to reduce variance, and treat theoretical parameters as context, not proof of experience. Before changing rank logic or an offer, use questions that expose real experimentation skill to test sample size, assignment, guardrails, and whether results are practical rather than random.
Short-window GGR needs the same caution: GGR = Bets - Wins, so an RTP dip can lift GGR without better catalogue quality. NGR is closer to decision value after bonuses, provider fees, payment costs, taxes, fraud, chargebacks, and relevant commissions.
Attribute catalogue effects before scaling
Begin attribution with decomposition:
Launches = Eligible visitors × Valid impression reach × Launch Rate
If launches rise, identify the component. More eligible visitors suggests traffic or availability; higher reach suggests placement coverage; higher Launch Rate suggests relevance, creative, rank, or intent. Post-launch depth tests whether added starts became meaningful play.
Where possible, randomise eligible users or sessions between treatments, log exposure, choose a primary outcome such as meaningful launches per eligible user, and guard against load failures, provider concentration, player choice, and NGR. Keep acquisition traffic, offers, and eligibility consistent between treatment and control.
Without randomisation, explicitly use weaker designs: matched cohorts by source, market, device, and lifecycle; difference-in-differences against an unaffected surface; and pre-change trends. Before-and-after charts cannot separate a release from campaign traffic, seasonality, provider outage, payment failure, or bonus change.
Data quality limits every ranker. Missing genres, duplicate game identities, stale availability, inconsistent provider names, and weak metadata undermine relevance; why personalisation starts with catalogue operations explains why these controls precede smarter ranking.
Instrument the decision path
Event data must reconstruct the journey. Impressions should record surface, module, rank, viewport status, game ID, provider, category, algorithm version, filter state, device, market, and eligibility outcome. Launches need timestamps and success or failure; gameplay needs successful load and the meaningful-play threshold.
Preserve catalogue snapshots. Historical analysis needs the provider label, category, availability, thumbnail, and rank rule seen by the player; otherwise analysts may compare yesterday's taxonomy with last month's exposure.
Audit denominators before interpreting movement: bot filtering, duplicate device identities, missing launch callbacks, unavailable titles counted as impressions, and preloaded-carousel exposures. Defects often resemble conversion problems until the event chain is inspected.
Build a scorecard around decisions
A scorecard should answer a decision. For weekly review, pair outcomes with diagnostics and guardrails:
- Meaningful launches per eligible visitor — primary evidence that the catalogue moved players into real play.
- Launch Rate by surface and rank band — separates relevance from exposure volume.
- Successful game load rate — identifies technical friction posing as weak demand.
- Session depth by comparable game type — tests post-launch quality without duration alone.
- Provider HHI and top-provider share — monitors concentration from merchandising.
- Freshness curve by game cohort — separates novelty from durable value.
- Cohort retention after meaningful play — tests whether experience supports return.
- NGR per eligible visitor — links outcomes to margin after direct costs.
Review by acquisition channel before claiming catalogue impact. Paid, affiliate, organic, CRM, and direct traffic differ in intent, device mix, and game familiarity. A rate improving only because high-intent traffic gained share has not earned catalogue credit.
The next catalogue review starts with falsification
Start with a falsifiable claim: “The new lobby order increases meaningful launches per eligible visitor for new mobile players without raising provider concentration or reducing seven-day retention.” Define population, exposure, conversion window, and guardrails before results.
Inspect eligibility, valid impressions, launches, successful loads, meaningful play, cohort return, and NGR in order. When a metric changes, first test traffic mix, availability, freshness, RTP variance, and tracking. A metric survives scrutiny when its denominator is stable, mechanism visible, segment result coherent, and claimed effect remains after plausible alternatives are challenged.