Retention curves: declining, flattening, smiling.

Two ecommerce brands can acquire customers at exactly the same first-month retention, and one is worth 50% more than the other. The difference is not in the first order, or even in the first drop-off. It is in the shape of the curve after that: whether it keeps falling, settles onto a plateau, or bends back up. Reading that shape is the single most useful thing a cohort chart tells you.

Most retention dashboards report a single number, and a single number cannot tell you which of those futures you are in. The curve can. This piece walks through what a retention curve is, why every one of them drops first, the three shapes it can settle into, and what each shape does to the two numbers you actually plan against: lifetime value, and the acquisition cost you can support.

What is a retention curve?

A retention curve, sometimes called a cohort survival curve, groups every customer by the period they first bought, then plots the share of that group still active in each following period. It is your retention rate drawn out over time rather than collapsed into one figure, and it is built with cohort analysis. The horizontal axis is time since first order; the vertical axis is the percentage of the original cohort still buying.

Because you follow one fixed group of customers, later movements are real behaviour and not new sign-ups papering over churn. A blended monthly active count can look flat while the underlying cohorts quietly bleed out, propped up by acquisition. The curve strips that illusion away. Every point on it refers to the same people you started with.

Why does every curve drop first?

Every cohort loses a large share of buyers immediately after the first purchase, and that first cliff is normal. First-time buyers are a mixture. Some were only ever going to buy once: a gift, a discount hunt, a one-off need. Others are the start of a genuine relationship. The initial drop is the one-time buyers leaving, and it says very little on its own.

This is why a low month-one repeat purchase rate does not by itself doom a cohort, and a high one does not guarantee it. The number that decides the outcome is what the survivors do after the cliff. A curve that keeps sliding and a curve that levels off can look identical for the first two periods and then diverge completely.

How do you read the three shapes?

After the first drop, a cohort curve settles into one of three broad shapes. The chart below draws all three from the same starting point so the divergence is visible.

• Declining (leaky). The curve never flattens and heads toward zero. There is no loyal core, so every dollar of revenue has to be re-bought with fresh acquisition. This shape caps lifetime value and leaves thin headroom on acquisition cost. It is common in commodity or one-time-need categories.
• Flattening (plateau). The curve falls, then levels onto a stable band. The plateau is your loyal core, the customers who have self-selected into a habit, and it behaves like an annuity: a predictable share who keep buying period after period. This is the healthiest common shape and the one that justifies paying up to acquire.
• Smiling (upturn). After the plateau, the tail bends back up because lapsed buyers reactivate faster than active ones churn. It is rare and strong, usually driven by winback, replenishment cycles, or a widening range that pulls dormant customers back along the customer lifecycle.

What does the shape do to LTV and to allowable CAC?

Lifetime value is the area under the curve, priced in margin. A flatter, higher tail is more area, and more area is more lifetime value to fund acquisition with. The quick geometric approximation most operators reach for assumes retention is constant, which is exactly the assumption the three shapes violate.

The worked example below runs the three curves from the chart through the same starting cohort and the same margin. Identical month-one retention, and the allowable acquisition cost swings by half.

Holding a 3:1 LTV to CAC ratio, the declining cohort supports about $20 of acquisition cost and the smiling one about $31, a 51% gap that opens entirely after the first order. Two teams reading only the month-one number would bid the same for a customer. One would be badly underpaying for a loyal base; the other would be overpaying for churn.

Why do curves flatten, and can you make one smile?

Flattening is not luck. It is arithmetic. In their paper on projecting customer retention, Peter Fader and Bruce Hardie showed that even if every individual customer has a constant, unchanging chance of lapsing, a cohort made of a mix of high- and low-loyalty customers will show a retention rate that rises over time. The low-loyalty customers leave first, so the survivors are increasingly the loyal ones, and the average climbs (Fader & Hardie, "How to Project Customer Retention," Journal of Interactive Marketing, 2007). A flattening curve is the expected signature of a customer base that contains a real loyal segment.

The smile is harder, and it is a lifecycle achievement rather than a statistical one. An upturn means your reactivation is outrunning erosion: winback flows, replenishment reminders, or a new range are pulling lapsed customers back faster than the plateau decays. You do not buy a smile with acquisition spend. You earn it on the existing base, which is why it is the most fundable shape a store can have.

Read your own curve before you read your ROAS. The dashboard tells you whether last week's spend paid back; the cohort curve tells you whether the customers you bought will still be here in six months, and therefore how much you were ever allowed to pay for them. Blufire builds these curves in contribution margin rather than revenue and reconciles the underlying orders to your ledger, so the area under the curve you are pricing acquisition against is real margin, not headline sales.