Optimization Metrics

When creating an optimization, you choose a target metric — the indicator that the ML algorithm will optimize — as well as secondary metrics for additional analysis.

Below is the complete list of available metrics, grouped by categories.

How to choose a target metric?

Choose the metric that directly reflects the business goal of the optimization. To increase overall revenue — ARPU. To improve retention — Retention D1 or D7. To optimize ad impressions — Ad Impressions per User. Add other metrics as secondary to see the full picture.

---

Revenue

Name	ID	Type	Description	Formula
ARPU — Average Revenue Per User	revenue_per_user	per user	Average revenue per user, including non-payers. Main KPI for evaluating overall monetization efficiency.	ARPU = Σ(revenue) / N_users
ARPPU — Average Revenue Per Paying User	revenue_per_paying_user	per user	Average revenue per paying user. Shows how much the audience that has already made a purchase spends.	ARPPU = Σ(revenue) / Σ(iap_conversions)
Revenue per Session	revenue_per_session	per session	Average revenue for one game session.	Revenue/Session = Σ(revenue) / Sessions
Revenue per Day	revenue_per_day	per day	Average revenue per day of user activity.	Revenue/Day = Σ(revenue) / Σ(days)
Revenue (Total)	revenue_total	total	Total revenue for the selected period. Used to evaluate overall financial performance.	Revenue = Σ(revenue)

---

Ad Impressions

Name	ID	Type	Description	Formula
Ad Impressions per User	ad_impressions_per_user	per user	Average number of ad views per user.	AdImpr/User = Σ(ad_impressions) / N_users
Ad Impressions per Session	ad_impressions_per_session	per session	Average number of ad views per session. Used to control the ad load on user experience.	AdImpr/Session = Σ(ad_impressions) / Σ(sessions)
Ad Impressions per Day	ad_impressions_per_day	per day	Average number of ad views per day of user activity.	AdImpr/Day = Σ(ad_impressions) / Σ(days)
Ad Impressions (Total)	ad_impressions_total	total	Total number of ad views for the period. Used to analyze ad volume.	AdImpr = Σ(ad_impressions)

---

Ad Revenue

Name	ID	Type	Description	Formula
Ad Revenue per User	ad_revenue_per_user	per user	Average ad revenue per user. Shows monetization efficiency via advertising.	AdRev/User = Σ(ad_revenue) / N_users
Ad Revenue per Session	ad_revenue_per_session	per session	Average ad revenue per session.	AdRev/Session = Σ(ad_revenue) / Σ(sessions)
Ad Revenue per Day	ad_revenue_per_day	per day	Average ad revenue per day of user activity.	AdRev/Day = Σ(ad_revenue) / Σ(days)
Ad Revenue per Impression	ad_revenue_per_impression	total	Average revenue from one ad view. Equivalent to eCPM.	AdRev/Impr = Σ(ad_revenue) / Σ(ad_impressions)
Ad Revenue (Total)	ad_revenue_total	total	Total ad revenue for the period. Shows the contribution of advertising to total income.	AdRev = Σ(ad_revenue)

---

Retention

Retention metrics show what proportion of users return to the game after a certain time. Available in two variants:

• Anytime — user returned at least once after the specified day. A softer criterion, suitable for evaluating overall loyalty.
• Calendar — user returned exactly on the specified day. A strict criterion, better for precise cohort comparison.

Retention (Anytime)

Name	ID	Type	Description	Formula
Retention D1	r1_rate	per user	Proportion of users who returned at least once after day 1.	D1 = Σ(re1) / N_users
Retention D3	r3_rate	per user	Proportion of users who returned at least once after day 3.	D3 = Σ(re3) / N_users
Retention D7	r7_rate	per user	Proportion of users who returned at least once after day 7. Allows evaluating long-term engagement.	D7 = Σ(re7) / N_users
Retention D30	r30_rate	per user	Proportion of users who returned at least once after day 30.	D30 = Σ(re30) / N_users

Retention (Calendar)

Name	ID	Type	Description	Formula
Retention D1 (Calendar)	r1cal_rate	per user	Proportion of users who returned exactly on the day following the first session. Evaluates onboarding quality.	D1 Cal = Σ(re_cal1) / N_users
Retention D7 (Calendar)	r7cal_rate	per user	Proportion of users who returned exactly on the 7th day. Evaluates content quality and medium-term engagement.	D7 Cal = Σ(re_cal7) / N_users
Retention D30 (Calendar)	r30cal_rate	per user	Proportion of users who returned exactly on the 30th day. Shows long-term product value.	D30 Cal = Σ(re_cal30) / N_users

---

Sessions

Name	ID	Type	Description	Formula
Sessions per User	sessions_per_user	per user	Average number of sessions per user.	Sessions/User = Σ(sessions) / N_users
Sessions per Day	sessions_per_day	per day	Average number of sessions per day of user activity.	Sessions/Day = Σ(sessions) / Σ(days)
Total Sessions	sessions_total	total	Total number of sessions for the period. Characterizes total volume of activity.	Sessions = Σ(sessions)

---

Engagement Time

Name	ID	Type	Description	Formula
Engagement Time per User	engagement_time_per_user	per user	Average time spent in the game per user (in minutes).	EngTime/User = Σ(time) / N_users
Session Time	engagement_time_per_session	per session	Average duration of one session (in minutes). Evaluates depth of immersion in the game.	SessionTime = Σ(time) / Σ(sessions)
Engagement Time per Day	engagement_time_per_day	per day	Average time in the game per day of user activity (in minutes).	EngTime/Day = Σ(time) / Σ(days)
Total Engagement Time	engagement_time_total	total	Total time of all users in the game. Characterizes overall audience engagement.	EngTime = Σ(time)
Engagement Days per User	engagement_days_per_user	per user	Average number of unique days the user visited the game.	EngDays/User = Σ(days) / N_users

---

In-App Purchases (IAP)

Name	ID	Type	Description	Formula
Transactions per User	iap_transactions_per_user	per user	Average number of purchases per user. Shows frequency of transactions.	Transactions/User = Σ(iap_transactions) / N_users
Transactions per Session	iap_transactions_per_session	per session	Average number of purchases per session.	Transactions/Session = Σ(iap_transactions) / Σ(sessions)
Transactions Total	iap_transactions_total	total	Total number of transactions for the period.	Transactions = Σ(iap_transactions)
FTP Conversion — First-Time Payer	iap_conversion_rate	per user	Proportion of users who made at least one purchase. Shows efficiency of the monetization funnel.	FTP = Σ(iap_conversions) / N_users
IAP Revenue per User	iap_revenue_per_user	per user	Average revenue from in-app purchases per user.	IAPRev/User = Σ(iap_revenue) / N_users
IAP Revenue per Session	iap_revenue_per_session	per session	Average revenue from in-app purchases per session.	IAPRev/Session = Σ(iap_revenue) / Σ(sessions)
IAP Revenue (Total)	iap_revenue_total	total	Total revenue from in-app purchases for the period.	IAPRev = Σ(iap_revenue)