Cohort-Based LTV:CAC vs Sitewide-Average LTV:CAC
Sitewide LTV:CAC looks healthy because legacy cohorts are subsidising the number. Here's when the cohort view tells the truth, when the blended view actually is honest, and how to report both.
Cohort-Based LTV:CAC vs Sitewide-Average LTV:CAC
Cohort LTV:CAC measures payback for a specific acquisition period; sitewide-average blends every customer you've ever acquired into one ratio.
Cohort-based LTV:CAC answers a specific question: for customers acquired in a defined window — say Q1 2024 — what's their measured contribution margin versus what you paid to get them? Sitewide-average LTV:CAC divides total historical contribution margin by total historical acquisition spend, blending every cohort you've ever run into one number.
The two rarely agree. When acquisition costs are rising and retention is softening — the current state for most Shopify brands in the €1M-€15M revenue band — the sitewide number lags the cohort number by 6-18 months. Finance tends to report sitewide because it reconciles to the P&L; growth teams need the cohort view because it reflects the CAC they're paying today.
The core problem: sitewide-average LTV:CAC is a weighted average dominated by whichever cohorts are largest and oldest. A brand that acquired 60% of its customer base in 2021-2022 — when Meta CPMs were half what they are now — will show a healthy 3.8x sitewide ratio while its 2024 cohort is actually paying back at 1.4x.
That gap isn't a rounding error. It's the difference between 'we're profitable' and 'every euro we spend today loses money on a 12-month horizon.' Both numbers can be technically correct and one of them is operationally useless.
Side-by-side: what each metric actually measures
| Dimension | Cohort-based LTV:CAC | Sitewide-average LTV:CAC |
|---|---|---|
| Question answered | Are customers we acquired in period X paying back? | Across all customers ever, what's the ratio? |
| Denominator | CAC paid in that specific window | Cumulative historical acquisition spend |
| Numerator | Contribution margin from that cohort only | Contribution margin from all customers ever |
| Lag before it moves | 0-3 months (real-time signal) | 12-24 months (heavily smoothed) |
| Reconciles to P&L? | No — needs cohort accounting overlay | Yes — matches finance's ledger view |
| Best for | Growth decisions, channel scaling, hypothesis testing | Board-level narrative, banking covenants, valuation |
| Fails when | Cohort is too small (<500 orders) or window too short | Legacy hero cohort or survivorship bias present |
| Typical DTC range (healthy) | 2.5x - 4.0x at M12 | 3.0x - 5.0x blended |
Notice the last row. The sitewide range looks better because it usually is better — it's flattered by every customer you acquired before CAC inflation. That's not the metric lying; it's the metric telling a truth about the past while your growth team is spending money in the present.
When each view is honest — and when it isn't
Cohort LTV:CAC is honest when the cohort is large enough to be statistically stable (rule of thumb: 500+ orders per cohort for a beauty or apparel brand) and when the observation window covers the second and third repeat purchase — typically M6 for consumables, M12 for apparel, M18 for higher-consideration categories like home goods.
It's misleading when you truncate the window too early. A subscription coffee brand looking at M3 cohort LTV:CAC will panic; the same cohort at M9 usually redeems itself. Picking the right cohort window matters more than which side of the debate you're on.
The hero-cohort effect
One outlier cohort — usually a viral moment, a hit product launch, or a pre-iOS 14.5 paid social window — can inflate sitewide LTV:CAC by 40% or more. If your 2021 cohort is 3x larger than every subsequent cohort and paid back at 5.2x, your sitewide number is telling you about 2021, not about the business you're running now. Segment the hero cohort out and re-run the blended math; the delta is usually sobering.
How to report both without a finance-vs-growth war
The reporting mistake is picking one. Finance needs the sitewide number because it ties to audited revenue and marketing spend on the P&L. Growth needs the cohort number because it's the only view that responds to a change in bidding strategy within the quarter. Report both, and be explicit about which question each answers.
The format that ends the argument is a quarterly cohort LTV:CAC waterfall: each column is an acquisition quarter, each row is months-since-acquisition, and the final row is a rolling sitewide blend. Finance sees the reconciliation, growth sees the trend, and the board sees both without needing to pick a side.
Cohort LTV:CAC (M12) vs sitewide-average LTV:CAC by quarter — typical DTC apparel brand
Cohort LTV:CAC (M12)
Sitewide-average LTV:CAC (rolling)
Frequently asked questions
Both, side by side. Lead with the cohort view for the most recent complete window (usually M12), then show the sitewide blend as reconciliation. If you only show sitewide, you're hiding new-cohort deterioration; if you only show cohort, finance can't tie it to the P&L.
At M12, 2.8x-3.5x is the operating range for apparel with 55-65% contribution margin. Consumables can sit lower (2.2x-2.8x) because payback happens faster; higher-AOV categories like furniture need 3.5x+ to absorb the return-and-refund tail.
Because it's a weighted average dominated by your largest and oldest cohorts. If half your revenue comes from customers acquired 2-3 years ago at cheaper CAC, their contribution keeps the blended ratio elevated even as new cohorts deteriorate. The lag is typically 12-24 months.
Rule of thumb: 500+ orders for stable margin math on an apparel or beauty brand, 1,000+ if you have wide AOV variance. Below that, one whale customer or one refund cluster distorts the ratio. Monthly cohorts are usually too small for €1M-€5M brands; roll up to quarterly.
Sitewide LTV:CAC only counts customers who exist in your database, which over-weights the ones who repeated at least once. Customers who bought once, churned silently, and never got refunded still count in CAC but drag less on LTV than they should. The effect inflates sitewide ratios by 10-25%.
Match the window to your repeat-purchase cycle. Consumables (skincare, coffee, supplements): M6 is meaningful because 2-3 repeats have happened. Apparel and accessories: M12 is the standard. Home, furniture, higher-consideration: M18 or M24. Reporting M3 for any DTC category will always look bad.
Report it both ways. Show the raw sitewide, then show it with the hero cohort excluded, and label the delta explicitly. Hiding the hero cohort is dishonest; not surfacing its distortion is also dishonest. Boards accept nuance if you show your work.
Yes — for a mature brand with stable CAC, stable retention, and no recent viral moment. If your CAC has drifted less than 15% over 24 months and cohort sizes are within 20% of each other, sitewide is a fair headline. Most brands in the €1M-€15M band do not meet those conditions.
Frame the cohort number as the leading indicator, not the verdict. Give the media buyer a target cohort LTV:CAC by channel (e.g. Meta prospecting needs 2.2x at M6) rather than pausing spend on a two-week signal. The goal is directional decisions, not panic pauses.
The sitewide number matches the P&L because both use cumulative totals. The cohort number won't reconcile directly — it's a different mathematical object. The reconciliation move is a cohort waterfall that sums back up to sitewide, so finance can see how the pieces roll up without changing their headline.
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.