Nine Numbers That Tell You a Slot Is Working

Integration · 2026-07-19 · 9 min read · By CROCO Games

GGR tells you what happened, not why it happened. Nine exposure-adjusted metrics for judging a slot, how to define each one, and the trap that makes each of them lie.

Most game-performance conversations start and end with GGR per title, sorted descending. It is a reasonable place to start and a terrible place to stop, because GGR is a product of three things — how many people saw the game, how many played it, and how much they lost — and the first term usually dominates. Sort by GGR and you will mostly rediscover your own merchandising decisions.

This is the set of numbers that actually separates a good game from a well-placed one, with the definition that makes each usable and the trap that makes it lie.

Why exposure adjustment is the whole game

The research foundation here is Auer and Griffiths' 2023 analysis of real operator data — 43,731 players and 763,490 sessions across five European countries — which found that a game's structural characteristics explain about 26% of the variance in games played per session but only 7.7% of the variance in bets per session. In other words: the game meaningfully shapes engagement and duration; the player brings the money. A KPI set built around monetary outcomes measures your traffic mix; a KPI set built around engagement measures the game.

Delfabbro's 2024 review of operator-data studies makes the complementary point: behavioural indicators drawn from real play reliably differentiate players and predict downstream outcomes, including account closure and self-exclusion. The signal is in behaviour, not in totals.

And because front-shelf placement is sticky, exposure differences between titles are large and persistent. Every metric below is therefore expressed per unit of exposure wherever possible.

The nine

1. Impression-to-play rate. Plays started ÷ tile impressions. This is the three-second layer's score — the thumbnail, name and badge doing their job. Trap: impressions must be counted as viewport impressions, not tiles rendered in the DOM; a lobby that lazy-renders 400 tiles will otherwise show a wildly deflated rate.

2. First-session length. Median rounds in a player's first session on the title. This is the clearest read on whether the game's early rhythm works. Trap: mixing bet sizes and volatility tiers — compare within a volatility band, not across the catalogue.

3. Feature-trigger rate in practice. Observed spins per feature entry, compared against the spec on the game sheet. A game whose live trigger rate diverges materially from its documented cadence is either misconfigured or misdocumented, and both are worth knowing before players tell you.

4. Return rate at day 7. Share of players whose first session included this title who played anything seven days later. Attributing casino-wide retention to a single game is imperfect, but the relative ordering across titles is informative. Trap: heavy promo cohorts inflate it — segment by acquisition source.

5. Repeat-play rate. Share of players who return to this specific title within 14 days of first playing it. This is the strongest single indicator that a game earned a place in someone's rotation, and it is where Hold & Win mechanics usually separate from the pack.

6. Rounds per exposure-day. Total rounds ÷ days the title held a merchandised position. This normalises the biggest confound — placement — and is the metric to bring to a shelf review. Trap: position within the shelf matters enormously, so band by position tier (1–3, 4–10, 11–30) before comparing.

7. Average bet relative to lobby average. Not to judge the game, but to identify who it attracts. A title pulling bets well above your lobby average is serving a different segment, per the player-segment playbook — that changes where it belongs, not whether it is good.

8. Session-to-session bet stability. Whether the players of a title escalate stakes over successive sessions. This is a dual-purpose metric: escalation is a commercial signal and a safer-gambling marker, and it belongs on both dashboards — see safer gambling by design.

9. Complaint and support-ticket rate per 10,000 rounds. Cheap to compute, rarely tracked per title, and an early warning for disconnect-recovery bugs, misleading promo copy or a mismatch between badge and experience.

What aggregator reports usually do not give you

Standard reporting arrives as rounds, turnover, GGR and unique players per game per period. That is enough for settlement and nearly useless for content decisions, because it lacks the denominator: impressions, position history, and per-player first-touch attribution. Three practical fixes:

• Log tile impressions and positions yourself. Position history is your data, it costs little to record, and without it every game comparison is confounded.
• Timestamp your merchandising changes. A dated log of shelf edits turns unexplained metric jumps into readable events.
• Ask providers for the spec that makes metric 3 meaningful. Documented feature cadence, hit frequency with the below-stake split, and RTP variant per market — see reading a PAR sheet.

Reading the numbers without fooling yourself

Three disciplines separate a working KPI practice from a dashboard hobby. Fix the comparison set — always compare a title to peers in the same volatility band and position tier, never to the lobby average. Respect sample size — a title with 4,000 rounds has a repeat-play rate with an error bar wide enough to swallow the difference you are excited about. Distrust single-metric verdicts — a game that wins on impression-to-play and loses on first-session length has a thumbnail writing cheques the math cannot cash, which is a merchandising fix, not a removal decision.

And when a metric moves, look for the merchandising event before you look for a player-behaviour explanation. Most step changes in per-title numbers are position changes wearing a disguise.

Frequently asked questions

What KPIs should operators use to evaluate slot games?

Exposure-adjusted engagement metrics rather than totals: impression-to-play rate, first-session length, feature-trigger rate versus spec, day-7 return rate, 14-day repeat play, rounds per exposure-day, bet relative to lobby average, bet stability across sessions, and complaints per 10,000 rounds.

Why is GGR per game a poor way to rank slots?

Because GGR is dominated by exposure. A title in position three accumulates turnover largely because it is in position three, so a GGR ranking mostly reproduces past merchandising decisions rather than revealing game quality.

Can a game really affect player retention?

Field data on 43,731 players indicates game structure explains roughly 26% of the variance in games played per session but under 8% of bet-size variance — so content meaningfully shapes engagement and session depth, while spend intensity is mostly brought by the player.

What data should an operator request from a game provider?

The documentation that makes live metrics interpretable: RTP variant per market, hit frequency split into full wins and below-stake returns, feature-trigger cadence, max-win probability, and the certification reference for the exact deployed build.

Key takeaways

• GGR per title mostly measures exposure; every useful game metric is expressed per impression, per session or per exposure-day.
• Field research shows game structure drives engagement far more than bet size, so engagement metrics measure the game while monetary metrics measure your traffic.
• Track nine numbers, and always compare within volatility band and position tier rather than against the lobby average.
• Aggregator reports lack the denominators that matter — log impressions, positions and merchandising changes yourself.
• When a metric jumps, check the merchandising log before inventing a behavioural explanation.

Partner with CROCO Games