Customer intelligence
In plain English
This screen answers: are the customers I'm getting now as good as the ones I got last year?
It uses a cohort table. That word just means "a group of customers who first bought in the same month." Everyone who first bought in January is the January group.
The table shows each group as a row, and then follows them forward: what percentage came back and bought again after 1 month, 2 months, 3 months, and so on.
Read the table straight down a column, not across. Here's why that matters. Pick the "month 3" column and read down it:
November group 38% came back by month 3
December group 34%
January group 29%
February group 24%
March group 21%
That's your business getting worse at keeping customers, month after month. And here's the dangerous part — your overall "repeat customer" percentage can look completely flat the whole time this is happening, because you're getting more new customers each month and the maths cancels out.
This table is the only place you'd see it.
Never compare across a row diagonally. A group that's only one month old will always look worse than one that's six months old — they've had less time to come back. Compare like with like: month 3 against month 3.
The second half of the screen shows which marketing channel your best customers come from, and how many months it takes before a customer has paid back what you spent getting them.
How to use it
- Click "Intelligence" in the left-hand menu, then "Customers."
- Find the "month 3" column in the cohort table.
- Read straight down it. Are the numbers going up or down as you move to more recent groups?
- If they're going down, your newer customers are worse than your older ones — usually because of heavy discounting or a change in where you're advertising.
- Scroll to the channel table. Look at "lifetime value" per channel, not just cost per customer.
- Check the "payback months" column before spending more on any channel. If it says 8 months, every pound you spend is tied up for 8 months — make sure you can afford that.
- Ignore any group with fewer than about 100 customers. Too small to mean anything.
⏱ ~10 min/month · 💳 Pro+ · 🎯 Know whether the customers you're buying are worth buying
Why this matters for your business
Aggregate customer metrics are the most reliably misleading numbers in commerce. "Repeat rate is 31%" feels like a fact about your business. It's actually an average across every cohort you've ever acquired, and averages hide exactly the thing you need to know: whether the customers you acquired this month are better or worse than the ones you acquired six months ago.
The failure is specific and common. Retention deteriorates steadily while the aggregate stays flat, because you're acquiring more customers each month and each new cohort retains slightly worse. The arithmetic cancels out. By the time the aggregate moves, you have six months of bad cohorts already in the base and a retention problem that takes a year to unwind.
The cohort matrix makes this impossible to miss. Each row is one acquisition month, tracked forward. Read down a column — say, month-3 retention — and you see the trend across cohorts directly. If November retained 38% at month 3 and March retained 21%, something in your acquisition changed, and you can see it four months before the aggregate does.
LTV by channel answers the companion question: which acquisition is producing the good cohorts. Not all customers cost the same and not all are worth the same, and a channel that looks cheap on CAC can be your worst channel once you know what its customers are worth.
What this typically unlocks
| Outcome | Typical result |
|---|---|
| Retention decline detection | ~4 months earlier than aggregate metrics |
| Channel decisions | Made on LTV, not on CAC alone |
| Cash-flow planning for acquisition | Payback months instead of a ratio |
| Hidden dilution from discount acquisition | Visible immediately in the matrix |
| Confidence in scaling a channel | High — you know what its customers are worth |
What you actually get
The cohort retention matrix
Customers are bucketed by the month of their first order. For each cohort, the matrix shows the percentage with at least one order in each subsequent month. Twelve months of cohorts by default.
| Cohort | Size | M1 | M2 | M3 | M4 | M5 | M6 |
|---|---|---|---|---|---|---|---|
| Nov | 840 | 24% | 19% | 38% | 16% | 14% | 13% |
| Dec | 1,220 | 22% | 17% | 34% | 15% | 13% | — |
| Jan | 1,410 | 19% | 15% | 29% | 13% | — | — |
| Feb | 1,380 | 17% | 13% | 24% | — | — | — |
| Mar | 1,290 | 15% | 12% | 21% | — | — | — |
Two ways to read it, and only one of them is useful most of the time:
- Across a row — how one cohort behaves as it ages. Useful for understanding your repurchase rhythm.
- Down a column — how successive cohorts compare at the same age. This is the one that tells you whether the business is improving. The M3 column above goes 38 → 34 → 29 → 24 → 21. That's a trend, and it's unambiguous.
Cohorts are compared at equal age, so a young cohort never looks bad simply for being young.
LTV by acquisition channel
| Column | What it means |
|---|---|
| Channel | How the customer was first acquired |
| Cohort size | Customers acquired through it |
| Mean LTV | Average lifetime spend |
| CAC payback (months) | How long until a customer repays their acquisition cost |
Payback is shown when acquisition cost is available for the same window. It uses the "ads only" spend basis described in the Revenue command center, so treat it as a floor — true payback is somewhat longer.
Why payback months beat an LTV:CAC ratio
An LTV:CAC ratio tells you whether a channel is profitable eventually. Payback months tell you whether you can afford it now, and that's usually the binding constraint.
| Channel | LTV | CAC | Ratio | Payback |
|---|---|---|---|---|
| Referral | $210 | $12 | 17.5:1 | 0.8 months |
| Organic | $186 | $0 | — | immediate |
| Paid social | $164 | $48 | 3.4:1 | 6.2 months |
| Discount/promo | $71 | $29 | 2.4:1 | 9.4 months |
Paid social at 3.4:1 is a perfectly healthy ratio. But a 6.2-month payback means every dollar of growth ties up cash for half a year. Scale it aggressively without the working capital and you can grow straight into insolvency — profitably, on paper.
Segments
The customer segments surface exposes the groups the platform maintains, including the ones that drive Action Center rules: dormant customers, at-risk customers, and high-value cohorts. Segment definitions and audience building live in Segments & cohorts; this surface is the analytical view of how those groups are performing.
How it works (without the technical bits)
Real merchant scenarios
Scenario A — The flat number hiding a collapse
Setup. Apparel brand. Repeat rate had read 31% for six months. Interpreted as "retention is stable, focus on acquisition."
The M3 column:
| Cohort | M3 retention |
|---|---|
| Nov | 38% |
| Dec | 34% |
| Jan | 29% |
| Feb | 24% |
| Mar | 21% |
Retention had almost halved. The aggregate stayed flat because each month brought more new customers than the last, and the mix arithmetic cancelled the decline.
Cause, confirmed by LTV by channel. A discount-led acquisition push had started in December. Discount-acquired customers had roughly 40% of the LTV of organic ones and barely repurchased.
Action. Rebalanced acquisition toward organic and referral, cut promo depth. Cohorts from the following quarter returned to 33% at M3.
How much earlier they knew. The aggregate repeat rate finally started falling in month seven. The matrix showed it in month three.
Scenario B — The cheap channel that wasn't
Setup. Merchant scaling a paid channel with a $22 CAC — by far their cheapest paid source. Plan was to triple spend.
LTV by channel:
| Channel | Size | Mean LTV | Payback |
|---|---|---|---|
| Referral | 640 | $224 | 0.7 mo |
| Organic | 2,100 | $198 | immediate |
| Paid — brand search | 890 | $176 | 3.1 mo |
| Paid — the cheap one | 1,450 | $58 | 8.9 mo |
The cheap channel had the worst customers by a wide margin. $58 LTV against a $22 CAC is a 2.6:1 ratio that takes nearly nine months to pay back.
What CAC alone had hidden. It looked like the best channel on the only metric they'd been watching.
Action. Held that channel flat and moved the incremental budget to brand search — 3× the CAC, 3× the LTV, and a payback period they could actually finance.
Scenario C — Payback as the real constraint
Setup. Fast-growing brand, healthy 4.1:1 LTV:CAC, scaling paid hard. Ran into a cash crunch despite growing 60% year over year.
Payback analysis:
Paid social CAC $48 LTV $196 ratio 4.1:1 payback 7.4 months
The arithmetic of the crunch. Every $48 spent came back over seven and a half months. Tripling monthly spend meant carrying roughly seven months of acquisition cost simultaneously. At $180,000/month of spend, that's over $1.2m of working capital tied up in customers who hadn't paid back yet.
Profitable and insolvent is a real state, and the ratio alone never shows it.
Action. Shifted mix toward referral (0.9-month payback) and capped paid growth at a rate their cash cycle could carry. Growth slowed to 40%; the business stopped being fragile.
Scenario D — A cohort that got better
Setup. Subscription business relaunched onboarding in February — a new welcome sequence and a first-order guide.
The M2 column:
| Cohort | M2 retention |
|---|---|
| Dec | 41% |
| Jan | 43% |
| Feb | 58% |
| Mar | 61% |
| Apr | 59% |
A clean step change at exactly the cohort the change shipped to, sustained across the next two.
Why the matrix proved it and a dashboard wouldn't. Aggregate retention rose only ~4 points over the same period, easily attributable to seasonality or noise. Cohort-level, the effect was unambiguous: it appeared exactly when expected, only in cohorts exposed to the change, and it held.
Result. The onboarding change was extended to reactivated customers too. Estimated annual LTV impact across the base: about $310,000.
Scenario E — Reading a young cohort correctly
Setup. Merchant panicked that their newest cohort showed 15% at M1 versus an older cohort's 38% at M3.
The error. Comparing different ages. M1 and M3 are not comparable numbers — customers accumulate repeat purchases over time, so a three-month-old cohort will always look better than a one-month-old one.
Correct comparison, M1 across cohorts:
| Cohort | M1 |
|---|---|
| Nov | 24% |
| Dec | 22% |
| Jan | 19% |
| Feb | 17% |
| Mar | 15% |
Still declining — but a 9-point decline, not the 23-point catastrophe the mis-read suggested.
Always compare down a column, never diagonally.
Best practices
✅ Read down the columns. It's the single most important habit for this surface.
✅ Pick one retention checkpoint and track it monthly. M3 works for most stores — long enough to be meaningful, short enough to be actionable.
✅ Check LTV by channel before scaling any acquisition source. CAC alone has misled more merchants than any other metric.
✅ Use payback months for budget decisions and the ratio for strategy. Cash cycle is usually the binding constraint.
✅ Annotate the matrix mentally against changes you shipped. A step change at a specific cohort is strong causal evidence.
✅ Watch discount-acquired cohorts specifically. They almost always retain worse; the question is by how much.
❌ Don't compare cohorts at different ages. Scenario E is the most common misreading.
❌ Don't trust a flat aggregate. It's the failure mode this whole surface exists to catch.
❌ Don't scale on ratio alone if cash is tight. Profitable growth can still be unaffordable growth.
❌ Don't over-read a cohort under ~100 customers. Small cohorts are noisy; the size column is there for this reason.
Plan tiers
| Capability | Free | Starter | Pro | Agency | Enterprise |
|---|---|---|---|---|---|
| Cohort retention matrix (12 months) | — | — | ✓ | ✓ | ✓ |
| LTV by acquisition channel | — | — | ✓ | ✓ | ✓ |
| CAC payback in months | — | — | ✓ | ✓ | ✓ |
| Segment performance view | — | — | ✓ | ✓ | ✓ |
| Dormant / at-risk segment feeds | — | — | ✓ | ✓ | ✓ |
| Cohort export | — | — | ✓ | ✓ | ✓ |
| Multi-store cohort roll-up | — | — | — | ✓ | ✓ |
Frequently asked
How far back do cohorts go? Twelve months of acquisition cohorts by default.
What counts as "retained" in month N? At least one order in that calendar month after acquisition. It's a binary — one order or ten both count as retained.
Why does month 3 spike for some businesses? That's your natural repurchase rhythm — consumables and subscriptions often show a clear cycle. It's a feature of your product, not an error.
Why is my newest cohort mostly blank? Those months haven't happened yet. Blanks are unknown, not zero.
Where does acquisition channel come from? Customer 360's acquisition attribution — the first touch credited to that customer. See Customer 360.
Why is a channel's LTV lower than my overall average? Because averages differ by channel — which is the entire point of the breakdown. Discount and marketplace channels typically sit well below organic and referral.
Is LTV predicted or actual? Actual mean lifetime spend to date. For forward-looking value, see Predictive LTV.
Why is payback missing for a channel? No attributable spend for that channel in the window — organic and referral typically have none, which is why their payback is immediate or not shown.
See also
- Growth Intelligence overview — the hub and its sections
- Revenue command center — KPIs and the CAC caveat
- Action Center — the rules driven by these segments
- Customer 360 — the per-customer view
- Segments & cohorts — building audiences from these groups
- Predictive LTV — forward-looking customer value
- Journeys & automations — acting on retention findings