Retention is the multiplier
Acquisition buys customers; retention decides what they are worth. A customer acquired at a given cost (CACCost to acquire one new customer.) contributes margin in each period they remain, so lifetime value (LTVTotal value a customer brings over time.) is the margin per period multiplied by the periods retained. Change the retention curve and every acquisition decision changes with it: the same channel, the same cost per customer and the same first-order margin can be a bargain or a loss depending on whether customers stay for two periods or twenty.
This is why a growth team spends so much of its time below the acquisition line. A small improvement in early retention is applied to every customer the business ever acquires, through every channel, which no acquisition optimisation can match. The tutorial on building a growth model in a spreadsheet carries the business's own retention curve forward cohort by cohort, so that a gain in early retention can be compared with an acquisition scenario in active customers a year out. It is also why retention is read from cohorts rather than from a blended rate: the blended rate mixes customers of every age, rises whenever acquisition grows, and hides whether newer customers are behaving better or worse. The cohort analysis tutorial builds the table; this guide is about reading what it produces and turning it into a payback figure. The growth economics hub sets both alongside the other unit economicsThe revenue and costs of one unit, an order or a customer, used to test if growth pays..
One distinction runs through everything that follows. "Retained" for a subscription means still paying; for a retailer it means ordering again within a window; for a product it means performing the core action. Pick the definition that matches how the business earns money and apply it consistently, or the curves will not be comparable with each other or with the acquisition cost they are judged against.
Curve shapes
Plotted with age along the bottom and share retained up the side, cohort curves fall into a few recognisable shapes, and each carries a different message.
| Shape | What it looks like | What it usually means | What to do |
|---|---|---|---|
| Flattening | A sharp early drop, then a plateau that holds | A retained core exists; customers who survive the early periods stay | Project lifetime from the plateau; aim activation work at the early drop |
| Steady decay | Falls at a roughly constant rate with no plateau | No retained core yet; the product or the fit is not holding anyone | Do not project a lifetime; fix retention before scaling acquisition |
| Cliff | Holds, then drops sharply at one age | A trial end, a price step, a contract term or a replenishment gap | Examine what happens at that age; test the transition |
| Smile | Falls, then rises in later periods | Seasonal or replenishment behaviour; customers return on their own cycle | Choose a grain that matches the cycle; measure "active in window" rather than "never lapsed" |
| Rising plateau | The plateau sits higher for newer cohorts | Product, onboarding or acquisition mix improving | Confirm the mix is stable before crediting the product; lock in what changed |
| Falling plateau | The plateau sits lower for newer cohorts | Acquisition mix drifting to worse-fit customers, often from saturating paid channels | Split by channel; expect rising marginal CAC alongside |
| Two-population | A plateau that is really an average of a loyal group and a one-off group | Blended curve hiding distinct segments | Segment until each curve has one shape; aim acquisition at the loyal segment's profile |
The flattening curve is the one every business wants, because a plateau is the only shape that justifies projecting lifetime value beyond the data. A steady decay means the honest lifetime value is bounded by the window observed, and a team projecting past it is guessing. The cliff is the most actionable, because something specific happens at that age and the experiment is obvious. The smile is the one most often misread, since a "never lapsed" definition turns a seasonal returner into a lost customer at the first gap.
Retention by segment and channel
A single retention curve for the whole business is an average of curves that may differ enormously, and the differences are where the decisions live.
By channel. The most important split, because channel CAC can only be judged against the retention of the customers that channel brought. It is common for the cheapest channel on cost per customer to produce the worst curve, and for an expensive channel such as organic search or referral to produce the best. A channel with low CAC and a steady-decay curve is a channel that acquires customers who do not stay, and no acquisition efficiency fixes that. Read channel curves at the same age and refresh them as cohorts mature, because channel quality drifts as campaigns, audiences and offers change.
By acquisition offer. Customers acquired on a deep discount, a free trial or a promotional bundle usually retain worse than customers acquired at full price, and the drop at the end of the offer produces a cliff. The offer may still be worth running if the retained share pays back, but only the split curve can say so.
By segment. Plan, product line, country, first-order value (AOVAverage revenue per order.), device, and whether a customer completed a particular onboarding action are the usual splits. The aim is to find the attributes that separate a loyal population from a one-off one, because that profile becomes the acquisition target and the activation goal.
By age at the time of a change. A pricing change, a product release or an outage shows up on the cohort table as a diagonal: every cohort shifts in the same calendar period. Reading the diagonal separates "newer customers are worse" from "everyone got worse in June".
The repeat purchase and loyalty guide covers the retail-specific mechanics of the second order and loyalty programmes; the lifecycle marketing guide covers the messaging that works on each stage of the curve. For subscription software, the SaaS industry guide shows why activation and churn, rather than sign-up volume, decide which channels a company can afford. The data design that makes channel curves possible, with analytics, billing and the CRM joined on one customer identifier, is set out in the guide to measurement for subscription businesses.
Payback on margin
Payback period is the number of periods until a customer's cumulative contribution margin equals the cost of acquiring them. It is a cash measure, and it has to be built on margin: revenue payback is a comfortable fiction that tells a business it has recovered its acquisition cost while the cost of goods, delivery and service it also paid are still unrecovered.
Payback period = the first period at which cumulative contribution margin per customer, read from the cohort's margin curve, equals or exceeds acquisition cost(periods)
Where margin per period is constant, payback is simply acquisition cost divided by margin per period. In practice margin per period falls as the cohort decays (fewer customers remain, so margin per original customer per period shrinks), so payback is read from the cumulative curve rather than computed from a single division. That is why the cohort table should be built a second time with cumulative contribution per original customer in the cells: the column in which that figure passes the channel's CAC is the payback period for that cohort and channel.
Three refinements matter. Include the ongoing costs of retention (loyalty rewards, service, retention offers) in the margin, because they are paid during the payback period. Use channel CAC, not blended, because payback is a channel decision. And where the curve is a steady decay, accept that payback may never arrive within the observed window, and say so, rather than extending the curve until it does. The break-even calculator gives the quick version, orders needed at a given margin to cover acquisition cost, before the full cumulative table exists, and the LTV and CAC guide covers the definitions of both inputs.
Payback read from a cumulative margin curve
A skincare subscription acquires customers through paid search at a channel CAC of £54, including the first-box discount. Contribution per subscriber per month is £13 while subscribed. The paid search cohort's retention curve reads 100%, 72%, 60%, 53%, 49%, 47%, 46% across months zero to six, then holds near 45%. Cumulative contribution per original customer is therefore £13.00 after month zero, £22.36 after month one, £30.16 after month two, £37.05 after month three, £43.42 after month four, £49.53 after month five and £55.51 after month six. Payback arrives in month six. On revenue (£29 a month) the same arithmetic would have claimed payback inside month two, which is the figure the previous agency had reported. The finance partner adds the loyalty credit paid from month four onward (£1.50 a month per active subscriber), which pushes payback to month seven, and that is the figure recorded in the unit economics table. Paid social, with a CAC of £41 but a steady-decay curve that reaches 22% by month six, has a cumulative contribution of £38.74 at month six and has not paid back; it is marked "no payback within window".
Cash constraints and growth speed
Payback matters because of cash. Every customer acquired is an outlay today against margin that arrives over the payback period, so a business growing quickly is always funding a widening gap between what it has spent on customers and what those customers have yet to return. The length of the payback period, the growth rate and the cash available together set a speed limit, and businesses that ignore it discover it as a cash crisis in the middle of their best quarter.
The logic is straightforward even without a model. If payback is short, the margin from recent cohorts funds the acquisition of the next ones and the business can grow from its own cash flow. If payback is long, growth has to be funded in advance, from reserves or investors, in proportion to how fast the business wants to grow, and a longer payback means more capital per unit of growth. Two businesses with the same lifetime value and the same CAC, one with payback in a few months and one in a couple of years, can afford very different growth rates on the same balance sheet.
This is also why payback, not the ratio, is the figure finance will ask about first, and why growth teams should keep both on the unit economics table. The ratio says whether customers are worth acquiring; payback says how many the business can afford to acquire this quarter. A plan that scales acquisition without checking the cash implication of the payback period is a plan to run out of money while profitable on paper. The practical settlement is a payback ceiling per channel, agreed with finance and derived from the cash position, beyond which spend is not increased however good the ratio looks. The tutorial on agreeing a target-led paid media budget uses that ceiling as the upper limit on spend that otherwise follows results. When the cohorts suggest money would do more keeping customers than buying new ones, the comparison of customer acquisition and retention sets out the signals for moving the split.
Improving retention and payback
Retention and payback improve through different levers, and knowing which is loose saves effort.
| Lever | Moves | Where it acts on the curve | Typical work |
|---|---|---|---|
| Activation and onboarding | Retention | The early drop | Reducing time to first value; removing setup friction; first-week prompts |
| Habit and frequency | Retention | The slope of the decay | Replenishment reminders, usage prompts, content that brings customers back |
| Offer and pricing transitions | Retention | The cliff | Softer trial endings; pause instead of cancel; plan fit |
| Lifecycle messaging | Retention | Across the curve | Stage-matched email and messaging, win-back for lapsed customers |
| Margin per period | Payback | The height of each step on the cumulative curve | Pricing, delivery and fulfilment cost, product mix, reducing retention incentives that do not retain |
| Acquisition cost | Payback | The threshold the curve must reach | Channel mix, creative, targeting, loop-produced customers lowering blended cost |
| Acquisition targeting | Both | Which curve new customers follow | Aiming spend at the profile whose cohorts flatten highest |
Work on the early drop pays most, because it applies to every cohort from now on and because a flattening curve is what makes the lifetime projection defensible. Margin work is often neglected by growth teams because it looks like finance, yet a modest increase in contribution per period shortens payback for every channel at once. And the acquisition targeting lever is the one that joins this guide to the previous one: once the segment curves are known, the most effective retention intervention is often to stop acquiring the customers who were never going to stay. For a service or a physical product, the early drop is worked on outside the inbox, and the tutorial on onboarding new customers beyond software removes the steps between purchase and first value.
The retention KPIA metric tied directly to an objective. on the growth dashboard should be a specific cell of the cohort table, such as retention at a chosen age for the latest complete cohort, and payback should be reported by channel from the cumulative margin table, refreshed monthly. Both will move slowly; both are worth more than any acquisition metric when they do.
The subscription walkthrough for the break-even calculator applies a retention curve and a chosen payback window to find the most a new subscriber can cost.
The retention objective hub turns this cohort view into a KPI set and shows which channels can credibly carry a retention plan.
For the channel that carries most retention work, email for retention sets lifecycle tactics and a budget justified by margin on incremental repeat orders measured against holdouts.
Trial length changes when revenue starts, and the free trial length test shows how to compare trial lengths without letting the shorter trial win simply by resolving first.
At the other end of the lifecycle, the cancellation save offer test measures whether a pause or discount offered to cancelling subscribers retains revenue net of its cost, with ease of cancelling as a hard guardrail.
Common mistakes
- Revenue payback. Costs of goods, delivery and service are paid during the payback period; only margin payback is real.
- Projecting a steady decay. Without a plateau, lifetime value beyond the observed window is a guess.
- Blended retention as the KPI. It rises whenever acquisition grows; define the KPI as a cell of the cohort table.
- One curve for all channels. Channel CAC judged against the business's average retention hides the channel that acquires customers who leave.
- Ignoring retention costs. Loyalty credits and retention offers are paid during payback and belong in the margin.
- Scaling on the ratio alone. A good ratio with a long payback can still run the business out of cash; agree a payback ceiling with finance.
- Misreading a smile. Seasonal returners counted as lapsed under a "never lapsed" definition make a healthy business look like a leaky one.
- Fixing the plateau before the drop. The early drop is where the lesson and the biggest gain are.
