Window must be stated — 30-day, 90-day, and 12-month RPR are not the same number.
The window has to be stated. A 30-day RPR, a 90-day RPR, and a 12-month RPR are not the same number for the same brand — they describe different stages of the customer relationship. Quote the window or the figure is meaningless.
Two formulations show up in dashboards, and they read very differently. The cohort form above ties the numerator and the denominator to the same group of customers acquired in a defined window. The looser period-based form — (customers with ≥2 orders in period) / (total customers active in period) — mixes new and tenured customers in the denominator, so the headline number drifts with whatever the new-customer mix happens to be that month. A heavy acquisition push can drop period-based RPR even when underlying retention is unchanged. The cohort form is the operator-useful read; the period form is at best a coarse health check.
RPR is the customer-level companion to AOV: AOV measures spend per transaction, RPR measures whether the customer returns at all. Together they sit inside a simplified LTV decomposition — roughly AOV × purchase frequency × gross margin. That framing is an operator read, not a canonical formula; LTV decompositions vary by accounting convention, contribution-margin definitions, and how segments are cut. What stays true across versions is that frequency is the slow signal. A 12-month LTV read requires twelve months of data; a 90-day cohort RPR is visible at day 90. That makes RPR the leading indicator operators actually get to use.
A concrete example makes the arithmetic testable against your own dashboard. Acquire 100 new customers in March. By day 90, 28 of them have placed a second order. The 90-day cohort RPR for the March cohort is 28 / 100 = 28%. Same brand, same customers, different window — the 30-day number for that cohort will be lower and the 12-month number will be higher.
What changes when RPR is low or high. Low cohort RPR shifts the case for retention investment — email, SMS, loyalty — because the acquired customer is not coming back on their own. Broken out by acquisition channel, low RPR reframes the channel as buying transactions rather than customers, which changes how the channel’s CAC should be read. Category matters too: consumables like skincare, supplements, or pet food often run meaningfully higher than considered-purchase categories like furniture or mattresses at the same window. Treat that as directional, not as a sourced multiplier, and benchmark within your own category rather than against the corpus.