Glossary

A controlled experiment that splits players into two or more variants to measure which produces a better outcome on a target metric.

Income earned from advertisements shown to players. Tracked separately from in-app purchase revenue so each channel can be optimized independently.

A behavioral cluster a player falls into, derived from their event history. Used for segmentation, content targeting, and offer tuning.

Revenue generated per active player on a single day. The standard daily monetization health indicator.

Revenue generated per paying player in a period. Isolates monetization performance from acquisition mix — if ARPPU rises while paying conversion stays flat, your spenders are spending more.

Revenue generated per active player in a period, paying or not. Captures monetization across the whole audience.

Mean duration of a play session. A primary signal of how engaging a session is once a player starts it.

A player who has stopped engaging with the game, typically defined as having no session for a fixed number of days (e.g. 7 or 14).

Percentage of active players who become inactive over a given period.

Ilara's ML-predicted probability (0–100) that a player will churn within a future window. Surfaced on every player profile and used to trigger retention interventions.

A group of players sharing a common attribute, most often install date or install week. Cohort analysis isolates how each new wave of players behaves over time.

Percentage of players from an install cohort who return to the game N days after install. The headline retention metrics for every live game.

Count of distinct players who had at least one session in a single day.

A composite 0–100 score capturing how actively a player interacts with the game across session frequency, depth, and event diversity.

The percentage of new installs who become paying users within N days of install. The conversion event is the player's first real-money payment; the window is anchored to install date. D1 = within 24 hours of install. D7 = within a week. This is the metric every D0/D1 offer engine is built to lift, and the target Ilara's first-user conversion model is trained on.

A real-money purchase made inside the game — a bundle, currency pack, item, or subscription.

Classification of where a player sits in their journey: new, active, at-risk, dormant, or churned. Drives most lifecycle automation in Ilara.

How much more often a model's top-N% scored players show the target behavior versus a random sample. "Lift @ top 1% = 30×" means the top 1% converts thirty times more than baseline.

Total revenue a player generates over their entire engagement with the game. "Predicted LTV" is Ilara's forecast of that value from early behavioral signals.

Count of distinct players who had at least one session in a calendar month.

Percentage of currently active players who have ever paid. Differs from paying conversion in that the denominator is active users, not an install cohort.

Percentage of users from an install cohort who have made at least one payment by a given point. D7 paying conversion is the most commonly reported version. For the install-relative variant focused on a player's first payment, see First-User Conversion.

A classifier quality score (0–1) summarizing the precision-recall curve. More informative than ROC-AUC when positive outcomes are rare, as with payer conversion.

Real-money income from in-app purchases over a period. Reported separately from ad revenue.

Revenue earned for every dollar spent acquiring a player. ROAS > 1 means a channel is profitable; the standard target horizon is D90 or D180.

A classifier quality score (0–1) measuring how well a model ranks positives above negatives. 0.5 is random; 1.0 is perfect ranking.

A defined group of players selected by shared attributes, behaviors, or model scores. The atomic unit for targeting experiments, offers, and notifications.

A continuous block of player engagement, from app open to inactivity timeout or app close.

One arm of an A/B test — either the control (current experience) or a treatment (the change being tested).

A high-spending player who generates disproportionate revenue. Typically the top 1–5% of payers by lifetime spend.