Purchase Frequency
Date: 2026-08-16
Orders per customer per period. It’s the third term in the value equation — alongside how many customers and how much each spends — and it’s the one with the most headroom, because nobody competes for it.
What it is
Purchase frequency is the average number of orders a customer places in a period.
On the running model:
15,000 orders ÷ 10,000 customers = 1.5 orders per customer per year
The value equation
Frequency is one of three multiplicative levers, and seeing them together is the point:
10,000 customers × 1.5 orders × £50 = £750,000
A 10% improvement in any one of them produces the same £75,000. But they differ enormously in cost:
| Lever | +10% costs |
|---|---|
| Customers | £22,500 of additional CAC, plus rising marginal cost — Marginal Analysis |
| AOV | Usually margin, via discounting or bundling — Discount Impact on Margin |
| Frequency | Lifecycle email and timing. Near-zero marginal cost |
Frequency is the cheapest of the three and the least worked on. That’s the argument for lifecycle investment stated in one table.
Frequency versus interval
The same information from two directions, and each suits different work:
frequency 1.5 orders/year → planning, forecasting, LTV
interval ~120 days median → campaign timing
Use frequency for the value equation and for Customer Lifetime Value. Use Time Between Orders when deciding when to send something.
Note frequency is a mean and the interval is best read as a median — the same distribution, summarised for different purposes, and the mean is the more fragile of the two.
Category sets the ceiling
Frequency is largely a property of what you sell, and pretending otherwise wastes effort:
consumables, replenishable 4–8× per year
apparel 1.5–3×
homeware 1–2×
furniture, appliances under 1×
For a low-frequency category, frequency is close to fixed and the leverage is elsewhere — AOV, margin, or referral. Setting a frequency target in a furniture business is a target nobody can hit.
For consumables it’s the opposite: replenishment timing is the highest-return work available, because the purchase is going to happen and the only question is where — Replenishment Timing.
Raising it
- Replenishment prompts timed to the product’s actual cycle, not a generic cadence
- Range expansion into adjacent categories, giving existing customers something else to buy. Usually cheaper than acquiring new customers for the same revenue
- Subscription or repeat-order options for consumables, which convert frequency from a marketing problem into a default — Subscription Metrics
- Post-purchase sequences in the window when propensity is highest — Lifecycle Messaging
- Loyalty mechanics, with the caveat that they often reward behaviour that would have happened anyway — Loyalty Programmes
Measuring it honestly
- Define “active”. Frequency depends entirely on who’s in the denominator. Customers who ordered in the period, or everyone ever acquired? The second falls every year automatically as the base ages
- Compute per cohort, or a large acquisition month drags it down with no behavioural change — Cohort Analysis
- Segment by category and by acquisition channel, since both drive it more than any campaign will
- It’s a mean over a skewed distribution. A minority of customers order far more often than the rest, so the average describes few of them — segment rather than optimise the mean — RFM Segmentation