Cohort payback: does a customer ever repay what you paid?

By the HonestTag team · Published July 12, 2026

Every acquisition budget rests on a claim about the future: this customer will eventually be worth more than they cost. Cohort payback is where that claim gets checked against what actually happened, month by month, with your own order data.

What the report is

A cohort is every identified customer whose first order landed in the same month. The March cohort is everyone who became a customer in March, whatever they bought and however they arrived. Guest checkouts, and returning customers whose first order came before HonestTag was installed, are left out, and the report says so. HonestTag tracks each cohort's cumulative ex-tax revenue per customer as the months pass, as placed (refunds are not subtracted from cohort curves; order edits and cancellations are), and draws the curve.

The payback marker is the point where that curve crosses what you paid to acquire the cohort: your blended NCAC over the trailing 90 days (connected-platform spend divided by new-customer orders). With a gross margin configured, the curve is contribution (revenue times margin) rather than revenue. Cross in month two and your customers repay their acquisition cost in two months. Cross in month ten and they take ten. And when a cohort at least three months old has never crossed, the report says "never cleared" in plain text instead of hiding the row or extending the axis until the problem scrolls out of view. The report is available on the Truth+ tier and up.

Why payback month matters more than NCAC alone

An NCAC above first-order contribution margin is not automatically a problem. Plenty of good businesses buy customers at a first-order loss. But that strategy is only financeable if repeat purchases repay the gap on a timeline you can actually fund. Payback month is that timeline, measured rather than hoped for.

A 60-day payback and a 12-month payback are different businesses at the same NCAC. The first recycles its ad budget six times a year from its own cash flow. The second needs working capital to bridge every cohort, and a bad quarter can stall the whole machine. Two stores with identical acquisition costs and identical average order values can sit on opposite sides of that line, and nothing on a spend dashboard will tell you which one you are.

The new cut: LTV by acquisition channel

Each cohort now splits by the first-click platform that acquired each member. The stamp happens once, at first-order time, from the recorded first click, and it is never revised or reconstructed later. If a customer's first recorded ad click was a Google click, they are a Google-acquired member of their cohort forever, no matter what they click next year.

This answers a question the first-order view cannot: which channel's customers actually come back. A channel can win the cheapest first order and lose the year. Another can look expensive at acquisition and quietly deliver the customers who reorder every quarter. The channel LTV curves put those trajectories on the same chart, at the same month index, from the same order data.

The 30-customer floor

Any cohort-by-channel cell with fewer than 30 customers shows unlock copy ("channel curves unlock at 30 customers; this cell has N") and no revenue number at all. Not a grayed-out estimate. No number.

This is deliberate. A per-customer curve built on 7 customers swings violently with one large order or one refund. A 7-customer LTV curve presented as truth is noise wearing a chart, and decisions made on it are decisions made on the shape of the noise. The floor is published, the count is shown, and the cell fills in on its own as the channel acquires its 30th customer.

The honest buckets, and the number we refuse to fake

Two groups get explicit buckets instead of being forced into a channel. Members acquired before channel tracking shipped render in a "(before channel tracking)" bucket: their first clicks are not honestly reconstructable, so no per-customer channel curve is claimed for them. And new customers with no recorded paid first click render as "(no paid first click)" rather than being smeared into whichever channel a model finds convenient. Every attribution HonestTag does record carries a proof record, so a customer sits in a channel because a specific click put them there.

One number is deliberately absent: per-channel payback. Marking a payback month for the Google slice would require a per-channel NCAC, meaning spend split by the same first-click rules as the customers, and that does not exist yet. So the blended payback marker stays on the parent report, and the channel view shows revenue curves without a crossing point. Printing one anyway would be fake precision, which is the thing this product exists to not do.

What the curve shapes show

1. Channels at the same month index. Whether a Google-acquired customer is worth more by month 6 than a Meta-acquired one is a direct read off the chart, a different question than which channel had the cheaper first order.
2. NCAC ceilings with repeat behavior priced in. A channel whose customers keep buying supports a higher NCAC than one whose customers vanish after order one. The same signal shapes how your verdicts read: a campaign that looks expensive on first-order math may be acquiring the best long-run customers.
3. Where the curves flatten. Curves that flatten early indicate growth has to come from new customers; curves that keep climbing indicate retention spend has real revenue to work with.

What the dishonest version looks like

Most attribution dashboards will happily show you channel LTV. What they rarely show is what is underneath it. Three patterns to check for anywhere you see an LTV chart: averages computed over tiny samples, presented at the same visual confidence as the big ones; a single blended LTV figure used to justify any acquisition cost on any channel; and channel LTV silently backfilled from a different attribution model than the one the rest of the tool reports, so the LTV page and the campaign page quietly disagree about who acquired whom.

How HonestTag computes it

Everything above runs on your store's own Shopify order data. Every order is classified new or returning against store order history; new customers attribute to the first ad click; every attribution stores a proof record. Refunds reduce store-level revenue metrics, while cohort curves remain gross of refunds. Google Ads and Microsoft Advertising receive the available retraction or restatement, and Klaviyo and Google Analytics 4 receive a refund event. HonestTag does not send Meta, TikTok, Pinterest, Snapchat, Reddit or OpenAI Ads a refund or retraction event; the refund is recorded in order proof. Every threshold in this report (the 30-customer floor, the bucket rules, the payback definition) is published in numbers on the in-app methodology page. Cohort payback and the channel cut sit on the Truth+ tier and up: correctness is never paywalled, breadth and judgment layers are. See cohorts on the product page for what the report looks like.

Frequently asked questions