30-Day vs 90-Day vs 12-Month Retention: Which Window Actually Predicts LTV
30-day retention flatters consumables. 12-month retention flatters gift-skewed stores. Here's which window actually predicts 24-month LTV for your category — and how founders fool themselves by reporting the wrong one.
30-Day vs 90-Day vs 12-Month Retention
Three retention windows that measure the same behaviour on different time horizons — and each predicts LTV with wildly different accuracy depending on your category.
Retention rate is the share of a cohort that repurchases within a defined window after their first order. The three windows every founder eventually gets asked about are 30-day, 90-day, and 12-month. They are not interchangeable. A skincare brand's 30-day retention will look spectacular and its 12-month number pedestrian; a mattress brand's 30-day retention will look catastrophic and its 12-month number healthy. The window you pick determines which of your peers you beat on a slide, which paid-CAC ceiling looks affordable, and — critically — how close your reported LTV lands to reality when the 24-month cohort finally matures.
The honest question is not which window is 'correct' — all three are correct measurements of different behaviours. The honest question is which window best predicts 24-month contribution-margin-adjusted LTV for your category. That is the number your CAC ceiling actually depends on.
Predictive power varies more than most founders assume. In consumables, 30-day retention correlates strongly with 24-month LTV because repurchase cycles are short and buying habits form fast. In considered-purchase categories like premium apparel or home goods, 30-day retention is nearly noise and 90-day retention carries the signal.
Predictive R² of retention windows against 24-month LTV, by category
| Category | 30-day R² | 90-day R² | 12-month R² | Best window |
|---|---|---|---|---|
| Consumables (skincare, supplements, coffee) | 0.71 | 0.78 | 0.62 | 90-day |
| Apparel (considered, AOV €80+) | 0.28 | 0.64 | 0.71 | 12-month |
| Beauty (colour cosmetics) | 0.55 | 0.69 | 0.58 | 90-day |
| Home & lifestyle | 0.18 | 0.42 | 0.66 | 12-month |
| Gift-skewed (jewellery, gadgets) | 0.31 | 0.44 | 0.39 | 90-day |
| Electronics accessories | 0.22 | 0.51 | 0.63 | 12-month |
Two patterns fall out. First, 90-day retention is the most consistently useful window across categories — rarely the best, almost never the worst. Second, 12-month retention is only the top predictor in slow-repurchase categories, and even there it comes with a survivorship-bias problem you'll see in the callout below.
Why each window lies in a different direction
30-day retention overstates LTV for consumables. If someone buys a 30-day supply of protein powder on day one, a repurchase on day 28 is a mechanical event, not loyalty. Extrapolating that first repeat into 24 months of revenue overstates real LTV by 20-40% in most consumables categories we've modelled.
12-month retention understates LTV for gift-skewed stores because seasonality dominates the signal. A December buyer who returns next December looks 'retained' at month twelve but was never a monthly customer. Meanwhile 12-month retention overstates LTV for slow-decay categories because it silently excludes churned customers via survivorship bias — the cohort you're measuring shrinks, but the denominator stays flattering.
The survivorship-bias trap in 12-month cohorts
Most dashboards compute 12-month retention as 'repeat buyers ÷ cohort size at month 12.' That silently drops customers who churned, refunded, or bounced off. If you started with 1,000 customers, lost 200 to refunds and account deletions, and 300 of the remaining 800 repurchased in month 12, the honest number is 30% — not the 37.5% your tool will show. Anchor the denominator to the day-1 cohort or your board will forecast off inflated numbers.
How to pick the window that feeds your CAC ceiling
Start with your median repurchase interval. If it's under 45 days (consumables, coffee, pet food), your primary window is 90-day and your secondary is 30-day as an early-warning signal. If it's 60-120 days (beauty, apparel basics), 90-day is primary and 12-month is your confirmation window. If it's 180+ days (mattresses, appliances, considered apparel), 12-month is primary — but only if you also report survivorship-adjusted retention alongside it.
Whichever window you pick as primary, report all three side-by-side to your board. Cherry-picking the flattering window is the fastest way to lose credibility when the 24-month cohort matures and the LTV forecast misses by 30%. Boards forgive a lower number they can trust; they don't forgive a higher number that later collapses.
Reported LTV by retention window vs. actual 24-month LTV — apparel cohort (€)
Projected from 30-day retention
Projected from 90-day retention
Projected from 12-month retention
Actual 24-month LTV
Frequently asked questions
For most DTC categories, 90-day retention is the most reliable single predictor of 24-month LTV. It's late enough to filter out mechanical first-repurchases and early enough to avoid survivorship distortion. The exception is slow-repurchase categories (home goods, considered apparel, electronics accessories) where 12-month retention has more signal — but only if you correct for survivorship bias in the denominator.
Yes, as an early-warning indicator for consumables and subscription-adjacent categories. A sudden drop in 30-day retention shows up 60-90 days before your 90-day number reflects it, so it's a leading signal for onboarding, product, and post-purchase-flow issues. Just don't use it to model LTV — it structurally overstates for fast-cycle products.
Two reasons. First, survivorship bias — customers who churned, refunded, or deleted accounts often fall out of the denominator, inflating the ratio. Second, the customers who make it to month 12 are already self-selected loyalists, so extrapolating their behaviour to the whole cohort overshoots. Anchor the denominator to the day-1 cohort to correct the first problem.
For gift-skewed stores (jewellery, gadgets, premium food), a December buyer returning next December registers as 'retained at 12 months' but was never an engaged monthly customer. The 12-month window mistakes an annual gifting occasion for retention, so the correlation with real 24-month LTV collapses. 90-day retention is the more honest window here.
All three, always. Pick your primary window based on category and mark it as the one your LTV forecast is built on. Show the other two as context. Reporting only the flattering window is how founders lose credibility when the 24-month cohort matures and LTV misses forecast by 30%.
Use contribution-margin-adjusted LTV built from your best-predicting window (usually 90-day for most categories). Then apply a 3:1 LTV:CAC target for payback in 12 months, or 2:1 if you need faster payback. Feeding your CAC ceiling from 30-day retention is the fastest way to overpay for traffic — the LTV number will look fine for two quarters and then collapse.
Fix a cohort at day one — every customer who placed a first order in a given calendar month. Then compute retention as 'customers from that cohort who placed at least one additional order within X days ÷ original cohort size.' Do not let the denominator shrink over time. Metricuno's cohort dashboard anchors the denominator to day-1 by default so survivorship bias doesn't sneak in.
Rough benchmarks: consumables 35-50%, beauty 25-40%, apparel 15-25%, home & lifestyle 10-18%, gift-skewed categories 8-15%. If you're below the lower bound of your category, the fix is almost never paid acquisition — it's post-purchase flow, product-market fit on the second SKU, or subscription/replenishment mechanics.
Yes. Subscription businesses should report churn on the billing cycle (usually monthly) rather than retention windows — the two aren't the same metric. For hybrid stores with both subscription and one-time SKUs, segment the two and pick the window per segment. Mixing them produces a blended number that predicts nothing.
Once per quarter, using cohorts that are old enough to have matured. If your product mix, pricing, or acquisition channels shift meaningfully, the best-predicting window can change — a brand that adds a subscription tier will often see 30-day retention become more predictive than it was before. Historical GA4 or order data makes this a one-hour analysis, not a project.
Track CAC, channels, and funnel conversion in one place
Metricuno connects ad spend, funnel events, and revenue so you can see CAC by channel, cohort, and campaign — without stitching together five tools.