Cohort Decay as a Leading Indicator: Spotting Sub-1 LTV:CAC Three Months Early
Your blended LTV:CAC still reads above 1, but the last three monthly cohorts are quietly decaying. Here's how to read the curves and act before the ratio flips.
Quick answer
Compare each new monthly cohort's month-2 repeat rate and month-6 cumulative LTV against the trailing 6-cohort median. When two consecutive cohorts land 10%+ below median on both metrics, your blended LTV:CAC will cross below 1 within roughly 90 days — even if today's ratio still looks fine.
Cohort decay as a leading indicator
Reading early cohort LTV curves to predict an LTV:CAC ratio collapse 60-90 days before it appears in blended numbers.
Blended LTV:CAC is a trailing metric. By the time the ratio drops below 1, the cohorts that caused it acquired customers months ago, and you've spent another quarter scaling on broken unit economics.
Cohort decay flips that. Each new acquisition cohort produces an LTV curve — cumulative revenue per customer over time — and the early shape of that curve (months 1, 2, and 6) reliably predicts where the full curve lands. When successive cohorts shift down in those early months, the blended ratio's trajectory is already set. You just can't see it yet.
The reader scenario: your monthly LTV:CAC report still shows 1.4 or 1.6. Finance is calm. But your last three Meta-acquired cohorts on Shopify are quietly underperforming, and nobody's read the curves yet.
Why blended LTV:CAC lags reality by a quarter
Blended LTV:CAC averages every active cohort. A six-month-old healthy cohort with strong repeat behaviour can mask three weak cohorts acquired since — for months.
The math is mechanical. If 70% of your contributing LTV comes from cohorts acquired 4-12 months ago, and those cohorts were healthy, a sharp decay in last month's cohort moves the blended number by maybe 3-5%. Not enough to trigger an alarm. By month three of decay, it's a 15-20% swing — and now you're below 1.
The 90-day blind spot
Most DTC finance teams review LTV:CAC monthly using a 12-month historical window. That window is what creates the blind spot: it takes roughly three fresh cohorts of decay to outweigh the healthy older ones in the average. Cohort curves remove the averaging — you see the new cohort's slope on day 30.
The two leading indicators to watch
Two signals from your cohort LTV curves predict the ratio break with usable lead time. Both are read cohort-over-cohort, not against an absolute target.
Signal 1 — Month-2 repeat rate: the share of a cohort that placed a second order within 60 days. Available 60 days after acquisition. Signal 2 — Month-6 cumulative LTV: total revenue per acquired customer at day 180. Available at day 180 but stabilises enough by day 90 to forecast reliably.
How to detect decay on your own curves
Pull the last 12 monthly cohorts and compute the trailing 6-cohort median for both signals. Plot each new cohort against that median. You're looking for two consecutive cohorts landing 10% or more below it on both metrics — that's the trigger.
A single weak cohort is noise — maybe a bad creative week, a promo cannibalisation, or a seasonal swing. Two in a row on both signals is structural. At that point you have roughly 60-90 days before blended LTV:CAC follows.
Worked example: apparel store on Shopify, monthly cohorts
| Cohort | Month-2 repeat rate | Month-6 cumulative LTV | Status vs 6-cohort median |
|---|---|---|---|
| Jan | 22% | €78 | Baseline |
| Feb | 21% | €76 | Baseline |
| Mar | 23% | €80 | Baseline |
| Apr | 22% | €77 | Baseline |
| May | 19% | €69 | −12% / −11% — first warning |
| Jun | 18% | €66 | −18% / −15% — trigger fires |
| Jul (blended ratio still 1.4) | 17% (forecast) | €63 (forecast) | Ratio projected <1 by Sep |
Pair this with driver isolation
The decay signal tells you the ratio is breaking. It doesn't tell you why. Once the trigger fires, run the CAC-inflation-vs-LTV-decay diagnosis to isolate whether your acquisition costs are creeping up or your retention is collapsing — the fixes diverge sharply.
What to do in the 90-day window
Freeze paid-spend scaling on the affected channels. The cohorts coming in right now are the ones that will drag the blended ratio under 1 — adding more of them accelerates the break. Hold spend flat on Meta and Google for the channels showing decay; reallocate test budget into retention experiments.
Then attack the month-2 repeat curve directly: post-purchase flow tightening in Klaviyo, replenishment prompts for consumable beauty SKUs, a second-order incentive sized below your unit margin. If decay is on month-6 LTV but month-2 holds, the issue is order frequency or AOV erosion — different fix, same urgency. The parent LTV:CAC diagnostic playbook walks through both branches.
Frequently asked questions
Reliably about 60-90 days before it shows in blended numbers. Month-2 repeat rate is available 60 days after a cohort lands; month-6 LTV stabilises enough by day 90 to forecast. Two consecutive weak cohorts on both signals is the threshold.
Month-2 repeat is the earliest signal that captures genuine retention rather than first-order behaviour. Month-6 LTV is where most DTC cohort curves have completed their steep early growth and the slope is predictive of the 12-24 month total. Together they cover both retention frequency and basket health.
Three to four cohorts is the minimum to compute a useful trailing median. If you're newer than that, importing historical order data from Shopify or GA4 into a cohort view buys you the baseline immediately. Without that baseline, you can't distinguish decay from normal variance.
For mature stores with stable seasonality, yes. For stores under €2M annual revenue or those with strong seasonal swings, widen the threshold to 15% and require three consecutive cohorts. Smaller cohorts mean noisier curves.
Yes — always. Blended cohort curves hide the channel doing the damage. Meta-acquired cohorts often decay first when ad fatigue or audience saturation hits, while organic and email cohorts stay healthy. Segmenting by channel gives you both the earlier signal and the action target.
Repeat purchase rate is a single number across all customers; cohort decay compares each acquisition cohort's curve to its predecessors. The cohort view catches a shift in newly-acquired customer quality that an aggregate repeat rate smooths over for months.
Less well. If your category has a natural 12-month repurchase cycle — premium electronics, mattresses — month-2 repeat is too sparse to read. Lean on month-6 cumulative LTV and AOV per cohort instead, and accept a 4-6 month lead time rather than 90 days.
Healthy apparel stores tend to land between 18-25% on month-2 repeat. Beauty and consumables run higher (25-40%). Home and electronics run lower (8-15%). The absolute number matters less than the cohort-over-cohort movement.
Yes. Compute the 6-cohort trailing median weekly, compare the newest fully-observed cohort, and alert when two consecutive cohorts breach −10% on both signals. Most analytics platforms can run that check on a scheduled query against your order data.
Blended LTV:CAC drops below 1 within roughly 90 days, and by then you've already acquired another quarter of weak cohorts at full spend. Recovery typically takes two quarters of frozen scaling plus active retention work — far costlier than acting on the leading signal.
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.