Rebuilt independently from the 15,091-row Stripe subscription export and reconciled against the audited 2025 P&L. The flagship $297 per 4 weeks plan churns hard at first renewal, yet acquisition is cheap enough that even a one-and-done buyer clears CAC. The 2025 loss came from growth experiments, not the core engine. The fresh 2026 data shows a deliberately smaller funnel with retention intact.
Only 9.3% of all subscriptions ever sold are still active. Retention splits sharply by interval: the weekly flagship is the leaky bucket and carries nearly all revenue; annual plans are the stickiest; most monthly subs are $0 comped accounts.
| Interval | Subs | Active | Active rate | Note |
|---|---|---|---|---|
| Weekly | 10,695 | 3.4% | leaky / ~all rev | |
| Monthly | 3,198 | 14.9% | 3089 bill $0 | |
| Annual | 1,198 | 47.3% | stickiest |
Survival of the $297/28-day flagship (10,695 subs). About half churn after a single 28-day cycle. Only 7.9% reach week 12 and 0.5% reach a full year. Median tenure is 2 weeks.
Percent still billing at week K. Log x-axis.
| Realized LTV (weekly) | Value |
|---|---|
| Median LTV (robust) | $594 |
| Mean LTV (tail-skewed) | $1,381 |
| Mean tenure | 4.7 wks |
Corrected per-subscriber CAC is ~$354 (the $112 figure is per paid acquisition including $19 trials). A single $297 charge nets ~$230 of contribution, so a one-and-done subscriber does NOT cover acquisition; it takes ~2 paid cycles to go green against a 2.77-cycle average life. LTV:CAC ~1.4x. Retention is a unit-economics problem, not just a growth problem.
New paid subscriptions per month. Green bars are 2026 (six months the 2025 P&L does not cover). Volume is down 41% year over year in H1, but the 4-week retention of each cohort has held in the 45 to 54% band. The funnel is smaller, not worse.
Grey = 2025, green = 2026. Hover for first-charge dollars. First-charge volume is an acquisition proxy, not collected revenue.
Acquisition volume halved year over year while retention quality stayed flat. This is consistent with cutting the growth experiments rather than demand collapse, but the 41% drop is the single biggest open question for diligence.
27 distinct price IDs across 11,995 paid subscriptions, but volume is concentrated in a handful. Duplicate IDs at the same price are prices recreated over time and should be consolidated.
| Price / interval | Subs | Share |
|---|---|---|
| $297/week | 9,561 | 79.7% |
| $197/week | 638 | 5.3% |
| $149/year | 452 | 3.8% |
| $249/year | 361 | 3.0% |
| $148/week | 189 | 1.6% |
| $297/year | 158 | 1.3% |
The Stripe export is a subscription roster with no charge history, so it cannot produce collected revenue. These figures come from the audited 2025 P&L and are the real topline. A normalized "MRR" off the active-sub snapshot overstates revenue by roughly 2x and is not used here.
| Gross revenue 2025 | $2,969,545 |
| Refunds & chargebacks | ($599,588) |
| Net revenue | $2,371,809 |
| Core engine EBITDA | +$505,664 |
| Growth experiments EBITDA | (983,084) |
| Total 2025 EBITDA | ($477,420) |
$599,588 of refunds and chargebacks on $2,969,545 gross is 20% of revenue. That is high and needs a direct explanation in diligence.
Strip the growth experiments and the core engine is +$505,664 EBITDA at 82% margin. The reported 2025 loss was the growth bucket, which returned almost nothing.
The CSV is authoritative for retention and acquisition. These three items close the remaining gaps and cannot be derived from it.
Computed from raw Stripe subscription export reconciled with the audited 2025 P&L. Active = current period end on or after Jun 23 2026 (Stripe snapshot proxy). First-charge dollars are an acquisition proxy, not collected revenue. Median LTV is the robust figure; the mean is skewed by a thin loyal tail. 2026 figures reflect data through Jun 15, 2026.