Second-order rate is the only retention metric that predicts anything

Why blended repeat rate drifts upward on its own

The standard retention metric on most DTC dashboards is some version of "percentage of customers who have ordered more than once." It is computed across the entire customer file, and it has a structural flaw that makes it close to unreadable: it moves for reasons that have nothing to do with retention.

Consider what happens when acquisition slows. Fewer new one-order customers enter the denominator, the file ages, more of the existing base has had time to place a second order — and the blended repeat rate rises. The brand celebrates improving retention during a quarter when it actually just stopped growing. The reverse happens during a strong acquisition quarter: a flood of new first-time buyers drags the blended number down, and a team that genuinely improved the customer experience gets told retention declined.

The metric is confounded by acquisition volume and by time. Any number that moves when your marketing budget moves is not measuring retention.

If your retention metric improves when acquisition slows down, it isn't a retention metric. It's a demographic report on the age of your customer file.

Cohort-locking fixes this. Take the customers who made a first purchase in a given month, follow only that group, and ask what share placed a second order within a fixed window. The denominator is frozen, the clock is identical for every cohort, and month-over-month movement now reflects something real. If you're measuring blended repeat rate today, the broader case for repeat purchase rate as a lever is worth reading alongside this — but the cohorted version below is the one to run your decisions on. This is the same discipline that separates a defensible lifetime value figure from a meaningless one — the fuller treatment is in how to calculate customer lifetime value.

The transition curve, and where the cliff is

Retention isn't a single rate, it's a sequence of conditional probabilities — and they are strikingly non-uniform. A representative shape for a consumable DTC brand:

• Order 1 → 2: 28% convert
• Order 2 → 3: 49%
• Order 3 → 4: 62%
• Order 4 → 5: 70%
• Order 5 → 6: 75%

The first transition is barely more than half as likely as the second, and it's well under half the fifth. This isn't a quirk — it's the most consistent pattern I see in DTC retention data. A first-time buyer has made one decision with limited information. A second-time buyer has validated the product, the shipping experience, and the brand. They're a different customer than they were six weeks ago, and the numbers treat them that way.

The practical consequence is that the leverage is enormously concentrated. Moving 3→4 from 62% to 65% touches a small, already-loyal population. Moving 1→2 from 28% to 31% touches every single customer you acquire, and it compounds through every subsequent transition, because a larger group crossing into the second order means a larger group entering the high-probability zone.

What one point of second-order rate is worth

Take a brand acquiring 4,000 new customers a month at $88 AOV and 41% contribution margin — $36.08 of contribution per order.

Model an average customer's remaining orders by chaining the transition rates above. Take 0.49, then 0.49 × 0.62, then × 0.70, then × 0.75, with the rate holding around 75% from there:

• Order 3 — 0.490
• Order 4 — 0.304
• Order 5 — 0.213
• Order 6 — 0.159
• Orders 7+ at a 75% plateau — 0.478
• Total: 1.64 additional orders beyond order two

So a customer who crosses the 1→2 threshold is worth approximately 2.64 orders beyond the first — call it $95.41 of additional contribution margin over their life. (Chain it yourself rather than taking the figure; the plateau assumption is doing real work here, and a brand whose later transitions settle at 65% rather than 75% gets a materially smaller number.)

Now move second-order rate by one point, from 28% to 29%. That's 40 additional customers per month crossing the threshold:

• 40 customers × $95.41 = $3,816 of additional lifetime contribution margin per monthly cohort
• Across 12 monthly cohorts: $45,800 per year of acquisition
• Three points — 28% to 31% — is $137,400 a year

And notice what this does to acquisition capacity. Lifetime contribution per acquired customer is the first order plus the probability-weighted value of crossing: $36.08 + (0.28 × $95.41) = $62.80. At 31% it's $65.66. That's about 5% more you can pay for a customer at the same payback period — and six points of second-order rate would buy you closer to 9%. In a market where CAC has been climbing for years, moving second-order rate is one of the few levers that expands what you can afford to bid — which is why retention and paid media are the same conversation, a point made at length in why your ROAS is lying about rising CAC.

Defining the cohort and the window

Three definitional choices determine whether the metric is usable, and getting any of them wrong quietly breaks the comparison.

Cohort by first-order date, not by any other event. Not by signup, not by first email, not by first session. The first purchase is the only unambiguous entry point, and it's the one every downstream probability hangs off.

Fix the window and never change it. Pick a horizon tied to your consumption cycle — 90 or 120 days for consumables, 180 to 365 for considered purchases — and hold it constant. A cohort measured at 120 days is not comparable to one measured at 90, and teams that let the window float end up comparing noise.

Exclude the incomplete cohorts. A cohort acquired 45 days ago cannot be evaluated on a 120-day window. Showing it on the same chart as mature cohorts creates a phantom decline at the right edge that panics people every single month. Either truncate the chart or plot immature cohorts on a separate partial-window basis.

One more refinement worth building: measure second-order rate on net customers, excluding those whose first order was fully returned. A customer who returned everything and never came back is not a retention failure; they're a returns problem, and mixing the two obscures both.

Reading a cohort at 60 days

The reason second-order rate is a forecasting metric rather than just a reporting one is that the 120-day figure is largely determined well before day 120.

Go back to the repurchase curve. If roughly 45% of eventual second orders arrive by day 60, then a cohort's day-60 second-order rate, divided by 0.45, gives a workable estimate of where it lands at day 120. A cohort sitting at 12% on day 60 is tracking to about 27%. One sitting at 9% is tracking to 20% — and you know that two months before the final number arrives.

That lead time is what makes the metric operationally valuable. It means a change to onboarding, packaging, or the post-purchase sequence gets a readable verdict in two months rather than four, and it means a bad acquisition cohort can be identified while you're still buying that traffic rather than after the quarter closes.

Calibrate the ratio from your own history rather than borrowing 0.45, and re-derive it annually. And treat it as a range, not a point — the estimate carries real error, and presenting it as a single number invites the same false precision that makes most forecasts undefendable.

Second-order rate by acquisition channel

The single highest-value cut of this metric is by acquisition channel, because it routinely inverts the ranking that CAC alone produces.

A channel delivering customers at $32 CAC with a 19% second-order rate and a channel delivering at $47 CAC with a 34% second-order rate look very different on a blended acquisition report — and the cheaper one usually wins the budget. Run the lifetime contribution math and the ordering frequently reverses, because the second channel's customers are worth substantially more over their life.

Discount-led acquisition is where this shows up most starkly. Customers acquired on a heavy first-order promotion reliably convert to a second order at lower rates than full-price customers, which means the CAC advantage of a promotional campaign is often borrowed from a retention cost that lands two quarters later. Segmenting second-order rate by whether the first order carried a discount is a one-query diagnostic that many brands have never run, and the answer tends to be uncomfortable.

What actually moves it

The interventions that move second-order rate are mostly not email campaigns. They cluster around the first experience:

• Getting the first product right. A customer whose first purchase was a poor fit for their need doesn't come back regardless of the flow. Merchandising toward a strong entry product beats optimizing the follow-up email.
• The delivery experience. Late or opaque first deliveries are among the most reliable predictors of non-repeat, and they're a fulfillment fix, not a marketing one.
• Usage and consumption support. Customers who never properly used the product have no reason to replace it. Content that drives real usage in the first two weeks moves the second order more than a reminder ever will.
• Timing the replenishment prompt to consumption, not to a calendar default — the same curve logic that governs win-back trigger timing.
• Not discounting the first order if it structurally lowers the second-order rate of the cohort it acquires.

Common questions

What is second-order rate in ecommerce?

Second-order rate is the percentage of first-time customers in a defined acquisition cohort who go on to place a second order within a fixed window. The cohort definition and the window are both essential. Measuring the share of all customers who have ever ordered twice produces a number that rises automatically as your customer base ages, which tells you nothing about whether retention is improving. A cohort-locked figure, such as the percentage of January's new customers who reordered within 120 days, is comparable month over month and actually actionable.

What second-order rate should a DTC brand expect in the first 120 days?

It varies enormously by category and consumption cycle. In the consumable DTC brands I have audited, a 120-day second-order rate below roughly 20 percent usually signals a product or first-experience problem, and the strongest performers run well above 40 percent. Those are working reference points from my own client work rather than published benchmarks, so treat them as orientation only. Considered and durable goods run far lower and should be judged on a longer window. The more useful comparison is almost always against your own trailing cohorts, because cohort-to-cohort movement isolates the effect of changes you actually made.

Why does the first-to-second order transition matter more than later ones?

Because it is where the largest drop-off occurs and where the conditional probabilities change most. The likelihood that a customer with one order places another is dramatically lower than the likelihood that a customer with two orders places a third, and higher again for a customer with three. Once a customer crosses into a second purchase, their probability of continuing rises sharply and stays relatively stable. That means effort spent moving customers from one order to two operates on the steep part of the curve, while effort spent moving three-time buyers to four operates on the flat part.

Build the cohort view

Most brands can produce a blended repeat rate instantly and cannot produce a cohorted second-order rate at all. That gap matters more than it sounds, because every retention decision — what to spend on lifecycle, which channel to scale, whether the onboarding change worked — depends on a number the dashboard doesn't have.

It's one query and one chart. The uncomfortable part is usually what it reveals about the channels currently winning the budget.