Retention Campaign Optimizer

Enter campaign parameters — the optimizer finds the targeting threshold that maximises ROI, calibrated on Two-Tower uplift model test-set results.

Model performance — test set

Metric	Two-Tower	Uplift LR	Random
Qini AUC	0.0187	0.0032	−0.0035
Uplift AUC	0.0129	0.0023	—
uplift@10%	2.41 pp	1.28 pp	—
uplift@30%	1.52 pp	0.61 pp	—

Optimal gift distribution — test set (284,718 users)

Your inputs

Total campaign budget (₽) —

Total money available to spend on retention offers

Average monthly revenue per active user (₽) —

How much a typical active user brings per month — from your analytics

Months remaining in the season —

LTV of a retained user: — · Offer mix: 91.3% unlock · 8.4% discount · 0.3% minutes · Avg. cost per offer: ₽45

Optimizer result — optimal targeting threshold

      Each percentile tier is included only if marginal uplift revenue (uplift × LTV) exceeds offer cost (₽45).
      Calibrated on Two-Tower test-set: Qini AUC=0.0187 · uplift@10%=2.41pp · uplift@30%=1.52pp.
    

Key performance metrics

Users targeted

—

Incremental retentions

—

Cost per retained user

—

Campaign ROI

—

ROI — Two-Tower vs Uplift LR vs random targeting