How to use Choosing The Right Cohort Window For Channel LTV Comparisons
A 90-day window makes Meta look like your best channel and Google Shopping look mediocre. Here's how to pick a cohort window that matches your repurchase cycle — and gives you an honest read on channel LTV.
Cohort Window For Channel LTV Comparisons
The lookback period over which you measure a customer's revenue when comparing lifetime value across acquisition channels.
A cohort window is the fixed observation period — 90, 180, 365 days, or longer — over which you accumulate revenue from customers acquired in a given month, then divide by cohort size to get channel LTV. The choice of window is not a cosmetic setting. A short window rewards channels that convert fast-moving buyers (Meta, TikTok) and penalises channels that acquire slower, higher-intent shoppers (Google Shopping, brand search, email). Pick the wrong window and you will over-invest in the channel that looks best on paper but underperforms on 12-month contribution margin.
Most Shopify and WooCommerce dashboards default to a 90-day LTV window because it's the shortest period that still captures a repeat purchase for most verticals. That default is the problem. Every channel has a different average time-to-second-order, and a 90-day window truncates the ones with longer cycles before they've had a chance to prove themselves.
If you allocate paid budget based on channel LTV — and most performance teams working with a €1M-€15M revenue base do — the window is arguably the single most consequential setting in your reporting stack. This guide walks through why the default is wrong, how to pick the right window for your category, and when to report multiple windows side by side instead of picking one.
Why the default window quietly distorts your channel mix
Meta and TikTok tend to acquire impulse buyers — someone scrolling, seeing a product, and buying within the same session or the same week. Those customers repeat quickly if they repeat at all. Google Shopping, brand search, and organic acquire buyers who have already been shopping around; they convert with higher intent but often after a longer consideration window and repurchase on a slower cadence tied to actual product need.
Inside a 90-day window, Meta cohorts get their full second-order revenue captured while Google Shopping cohorts have often only just placed their first order. The result: Meta LTV looks 20-40% higher than Google Shopping LTV. Extend the window to 365 days and the ranking frequently flips, because Google Shopping customers keep buying past day 90 while Meta customers plateau.
This is the mechanic behind the classic reallocation mistake: a performance team sees Meta beating Google Shopping on 90-day LTV, shifts €30k of monthly spend from Shopping to Meta, and watches blended CAC climb over the next two quarters. The channels weren't being compared fairly — they were being compared on mismatched cohort maturity.
The 90-day trap
If your finance team reports 90-day channel LTV and your paid team optimises against it, you are systematically over-funding fast-repeat channels and under-funding slow-repeat ones. The bias compounds every quarter.
The anchor: median time-to-second-order
The honest way to pick a window is to anchor it to your own repurchase behaviour, not a round number. The metric to compute is the median time-to-second-order for each channel — the number of days between a customer's first and second purchase, measured at the 50th percentile so outliers don't skew it.
Take the slowest channel's median time-to-second-order and multiply by roughly 2x. That's your minimum defensible window. If Google Shopping customers place their second order at a median of 140 days, a 280-day window is the floor for a fair comparison. Anything shorter and you're cutting off the slower cohort before it matures.
Channel LTV at 90, 180, and 365 days — same customers, different story
90-day window
180-day window
365-day window
Notice how Meta leads at 90 days, ties Google Shopping at 180 days, and trails it materially by day 365. Every real Shopify apparel or beauty account we've audited shows a pattern in this shape — the crossover point varies, but the direction is consistent. If you only ever look at the 90-day column, you never see it.
Category-specific rules: skincare, apparel, homewares
Median repurchase cycles cluster tightly by category, which lets you skip the analysis for a first-pass window choice. Skincare and consumables run on a 60-90 day cycle because the product physically runs out. Apparel runs 120-180 days and is heavily distorted by seasonality — a spring cohort's second order often lands in the autumn drop. Homewares, furniture, and considered categories need 270+ days because the second purchase is a separate need, not a replenishment.
The category rule is a starting point, not a substitute for measuring your own cohorts. A skincare brand with a strong subscription program will run tighter than 60 days; a fashion brand with a mono-drop model may need to move to a 365-day window to escape the seasonal trap. Use the category default until you have 12 months of clean channel-tagged order data, then switch to your measured median.
Recommended minimum cohort window by category (channel-LTV comparison use case)
| Category | Median time-to-2nd order | Minimum window | Preferred window | Why |
|---|---|---|---|---|
| Skincare / beauty consumables | 35-50 days | 90 days | 120 days | Product runs out; repeat is replenishment-driven |
| Supplements & vitamins | 30-45 days | 90 days | 120 days | Monthly consumption cycle |
| Apparel (multi-drop) | 70-110 days | 180 days | 240 days | Repeat tied to next drop / season |
| Apparel (mono-drop / seasonal) | 140-200 days | 270 days | 365 days | Second order often lands next season |
| Footwear & accessories | 120-180 days | 270 days | 365 days | Lower purchase frequency per SKU |
| Homewares & décor | 180-260 days | 365 days | 540 days | Second purchase is a new need, not replenishment |
| Furniture | 300-500 days | 540 days | 730 days | Considered category; long inter-purchase interval |
| Electronics & gadgets | 220-320 days | 365 days | 540 days | High AOV, low frequency |
If you sit at the edge of two categories — say a skincare brand with a device SKU, or a homewares brand with a candle refill line — split your reporting. Run a short window for the consumable line and a long window for the considered line. Blending them under one window hides which product mix each channel is actually acquiring.
Rolling vs fixed windows, and when to report multiple
Once you've picked the right window length, the second choice is whether to run it as a rolling window (the last 180 days from today) or a fixed cohort window (customers acquired in Jan 2024, tracked to Jul 2024). Rolling is easier to slot into monthly reporting; fixed cohorts are truer for year-over-year channel comparison because seasonality and promo calendars align.
The strongest reporting setup shows both. A rolling 90-day view for weekly paid-media decisions, a fixed 365-day cohort view for quarterly budget allocation. If you can only afford one, pick fixed cohorts at your category's preferred window — that's the number that should govern where next quarter's money goes.
When to report multiple windows side by side
If a channel's LTV ranking flips between your 90-day and 365-day view — which is common for Google Shopping, brand search, and email — publish both windows in the same report. Forcing stakeholders to see the gap prevents someone downstream from quoting the more flattering number in isolation.
Frequently asked questions
Meta's targeting and creative format acquire fast-decision buyers whose repeat behaviour concentrates in the first 60-90 days. Inside that window their cohort is fully mature, while slower channels like Google Shopping and brand search are still ramping. Compare them at 90 days and you're comparing a finished cohort to a half-baked one.
Pull all customers acquired in a fixed month at least 12 months ago, tag each with their first-touch channel, then compute the days between order 1 and order 2 for those who repeated. Take the 50th percentile per channel. If you're on Shopify with a channel-tagged order export or Metricuno's channel LTV view, this is a single query.
First-touch. Last-touch bleeds email and direct into every cohort and makes paid channels look worse than they are. LTV comparisons are about which channel found the customer, so first-touch is the honest attribution model for this specific use case.
Yes, but with a twist. TikTok tends to acquire even faster-repeat behaviour than Meta in the first 30 days, then falls off harder past day 120. The 90-day window flatters TikTok even more than Meta, which is why TikTok-acquired customers often show lower 12-month repeat rates despite strong 90-day LTV.
Use a proxy: assume the new channel behaves like the closest existing channel with a full history (TikTok as proxy for Reels, Pinterest as proxy for Meta prospecting, etc.), then apply that channel's mature cohort curve. Revisit the assumption every quarter as your own data accumulates.
Even more. CAC is fixed at acquisition, but LTV grows with window length. A CAC:LTV computed on a 90-day window versus a 365-day window can differ by 2-3x for the same channel, which means the same channel can look unprofitable or profitable depending purely on your window choice.
A spring cohort's second order often lands in the autumn drop — 150-200 days later. If your window is 120 days, you cut them off before their next purchase. Apparel windows should always span at least one full seasonal cycle, which is why 240-365 days is the honest range for multi-drop brands.
Predicted LTV is useful for forward-looking allocation, but the model itself needs to be trained on actual cohort observations at a sensible window. Garbage in, garbage out: a pLTV model fit on 90-day data will systematically under-predict slower channels' true value. Use pLTV to project past your observation window, not to skip it.
Once a year, or whenever your product mix shifts materially (new category launch, subscription program, price repositioning). Median repurchase cycles are stable within a category and product mix but change when you fundamentally change what you sell.
Keep the 90-day payback as a cash-flow constraint, but separate it from channel comparison. Report 90-day contribution margin (for cash discipline) and 365-day LTV (for channel allocation) as two distinct numbers. Conflating them is what forces the misallocation in the first place.
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.