Click an ad set to expand its ads. Last 7 days, by event date.KPI band Subs / Revenue / CPA = cohort view: trials STARTED in the window, everything they have paid to date (renews + grace included, Revenue net of Apple 15%, QA sandbox filtered). v2: per-ad cohorts built from our own Adapty profile-events feed. Pending = trials still running now (window trials minus window cancels and window payments — a 3-day trial does convert inside 7 days). Billing issue = 0 here — it is a >7d lifetime state, not a window metric. Trial→paid (−4d) shows — : payments cannot be linked to their start cohort from daily aggregates. Reach / Frequency = Meta-deduplicated across the window (true window frequency). — Unmapped = clicks where Meta failed to fill {{ad.id}} in the link — real ad traffic, unattributable to an ad. Projected CPA = spend / ((paid + Σ age-weighted active trials + open Grace × 30%) × 100%); age weights depend on 3d/7d duration. Attributed +30% = one shared unattributed baseline allocated across Main, Agency and Spanish by attributed Projected Revenue; source totals add back to All; Organic uplift ≈ = unattributed Projected Revenue above that baseline, with the same share applied to cohort metrics.
Legend. Revenue = net of Apple 15%; everything cohorts paid to date (first payments + renewals + grace recoveries), refunds subtracted. Projected Revenue = actual net revenue received + each pending user’s vendor-product net × its 3d/7d age weight + open Grace net × 30%, then the projected part × non-refund probability. ROAS = revenue ÷ spend; Projected ROAS = projected revenue ÷ spend. CPA = spend ÷ paid subs. Projected CPA = spend ÷ ((paid + Σ age-weighted active trials + open Grace × 30%) × 100%). Active 3d/7d weights change with age; ×100% = minus refunds (0.0% so far). Attributed +30% takes one shared baseline from the unattributed bucket and allocates it across Main, Agency and Spanish by attributed Projected Revenue; source totals add back to All. Organic uplift ≈ is the remaining unattributed Projected Revenue; its share of the unattributed bucket estimates the other cohort metrics. It is part of the total, not extra revenue. Trials / Pending / Cancelled — крупное число = все триалы вместе; подпись «3d · 7d» — сплит по длине триала (7d = winback-оффер после закрытия пейвола). Pending = ни отмены, ни оплаты, первое списание ещё не наступило (по реальной дате окончания 3d/7d/30d-триала). 30d trial = 30-дневный intro-оффер основного пейвола (с 23.09): участвует во всём наравне с 3d/7d, включая Projected CPA; его реальная конверсия откалибруется на первых списаниях ~23-24.10. Cancelled = auto-renew off в триале. Stayed = (trials − cancels) ÷ trials. Grace = billing retry in progress. Billing dead = no charge, never recovered.