| Segment | Players | % Total | Strategy | Notes |
|---|
level_start; measured drift vs the old definition: Hexabug 1%, Hole Em All 4% avg / 5% p90 over 3,000 levels with n≥100). "IAP user" = purchased at least once up to the end of the selected active-date window. Hexabug: coin from IAP is inferred as purchases × pack size — the event cannot separate it. Hexabug: numbers are roughly half the old tab's, because the ref_value coin-conversion rows fired alongside every real event are now excluded (they double-counted). Hole Em All: heart_time / battery_time / clock_time are minutes of regeneration, not counts. Hole Em All: campaign / ad network only complete from 09/06/2026. Starting grant per user (never logged as a source event) — Hexabug: 900 coin · 5 heart · 1 per booster · Hole Em All: 0 coin · 5 heart · 1 per booster (battery, clock, compass, freeze, growth, magnet).
| Metric | Segment | Control | Variant | % Change | p-value | q (FDR) | P₊ | MDE achieved | CI95 | Sig |
|---|
Each row is one action; the columns are how many times that same action repeats — time 1, time 2, time 3… This is not a multi-step funnel but an adoption-depth funnel: it answers "how many people tried it, and how many came back for more". In Distinct days mode, the "2 days" column means the action happened on at least 2 different days (not necessarily consecutive — could be D0 and D5), not "on day 2".
The cohort only contains fully mature users (installed at least one full observation window before the last complete day), so no cohort is cut short in the right-hand columns; the source table's last day is dropped because it is always partial. Every number is recomputed from the underlying rows when filters change — including the eligible denominator and the level / lost / won state: narrowing the days since install window narrows that state too, so it never carries information from the future.
lens/shared/lenses/level.js: Heatmap × Week · Level Grid · Compare 2 weeks · A/B mode + CSV. In the meantime, the original version is available at /lens/ (select Hexabug → tab “Heatmap Level”).grey band = Play p10→p90 · dot = hb
cohorts mature for 7+ days only
DAU · installs · first-time buyers
hashed band = US table benchmark, by platform
BigQuery has no ad spend ⇒ inferred from revenue
all ARPI/ROAS above use GREATEST per (user, day)