Decomposing Revenue by New vs Returning Customer Cohorts
How to decompose revenue by new vs returning customer cohorts so you know whether AOV moved because of acquisition mix, checkout UX, or repeat-buyer behavior.
Quick answer
Split AOV × Orders into two parallel equations — one for new customers, one for returning — then compare period over period. If new-cohort AOV moved, it's an acquisition-mix or landing-page problem (CRO). If returning-cohort AOV moved, it's a retention, bundling, or lifecycle problem. A blended AOV change without this split is almost always misread.
Decomposing Revenue by New vs Returning Customer Cohorts
Splitting the AOV × Orders revenue decomposition by acquisition cohort to isolate whether new or returning buyers drove the change.
Cohort revenue decomposition takes the standard Revenue = AOV × Orders identity and runs it twice: once for first-time buyers in the period, once for customers with at least one prior order. The result is four moving parts — new AOV, new orders, returning AOV, returning orders — instead of two blended numbers.
This matters because a flat blended AOV often hides a 15% drop in new-customer AOV being masked by a returning-cohort spike, or vice versa. The split tells you whether the next experiment belongs in the acquisition funnel, the checkout, or the post-purchase lifecycle.
The parent framework — AOV × Orders revenue decomposition — answers the question "did revenue move because of basket size or order count?". It cannot answer "which kind of customer caused it?". That second question is usually the one your CFO actually asked.
On a Shopify store doing €3M, a 4% blended AOV decline reads as background noise. Split by cohort, the same 4% can be a 12% drop in new-buyer AOV — a clear signal that a paid-social campaign is pulling lower-intent traffic into the funnel.
Why blended AOV hides the real story
New and returning customers behave like two different stores sharing a checkout. New buyers test a single hero SKU, use discount codes, and convert from paid channels. Returning buyers add 2-3 items, use saved payment methods, and arrive from email or direct.
When you average them together, the AOV you see is a weighted mean of these two populations — and the weighting itself shifts every month. A Black Friday week tilts the mix toward new buyers; a January re-engagement campaign tilts it toward returning. The AOV moves even if neither cohort's behavior changed.
The mix-shift trap
If your new:returning order ratio changes by more than 5 percentage points between two periods, any blended-AOV interpretation is unreliable. Always check the ratio before drawing conclusions from a top-line AOV chart.
How to run the decomposition
Pick a period (usually month). For each order in the period, classify the buyer as new (zero prior orders before the period start) or returning (one or more prior orders). Sum revenue and count orders within each group.
Compute four numbers per period: new_AOV, new_orders, returning_AOV, returning_orders. Repeat for the comparison period. The percentage change in each of the four is your diagnostic — the largest mover is your headline.
On Shopify, the customer.orders_count field gives you the classification directly. In GA4 you can approximate with the new_vs_returning user dimension joined to purchase events, though session-based attribution will misclassify cross-device buyers — Shopify-side data is cleaner.
Typical cohort splits by vertical
Indicative new vs returning AOV and order-share benchmarks by vertical (€1M-€15M DTC)
| Vertical | New AOV | Returning AOV | Returning AOV uplift | Returning share of orders |
|---|---|---|---|---|
| Apparel | €68 | €92 | +35% | 38% |
| Beauty & skincare | €42 | €71 | +69% | 55% |
| Consumer electronics accessories | €55 | €78 | +42% | 22% |
| Home & lifestyle | €85 | €118 | +39% | 31% |
| Food & supplements | €38 | €64 | +68% | 62% |
Returning AOV runs 35-70% above new AOV in most consumable verticals because repeat buyers stack subscriptions, add complementary SKUs, and qualify for free-shipping thresholds with fewer items in front of mind. If your returning uplift is below 20%, your post-purchase merchandising is the lever — not new-customer acquisition.
Reading the four-number diagnostic
If new AOV fell and new orders rose: cheaper-intent traffic — usually a paid-social audience expansion or a discount-led campaign. Fix in the acquisition funnel: tighten audiences, test landing pages that pre-frame value over price.
If returning AOV fell with stable returning orders: your repeat buyers are buying the same SKU but skipping the add-ons. Check whether you removed a post-purchase upsell, changed email recommendations, or hit a stock-out on a common cross-sell pair.
Experiments worth running once you know which side moved
New-cohort AOV is down: test a free-shipping threshold set 15-20% above current new AOV, a hero-bundle PDP variant, and a checkout-page recommendation widget. Each lifts new AOV without touching acquisition spend.
Returning-cohort AOV is down: test a logged-in homepage with a curated "buy it again + try this" rail, a subscribe-and-save prompt on the second-order confirmation, and a tiered loyalty threshold sitting just above the current returning AOV. These move the lifecycle, not the funnel.
Frequently asked questions
The parent decomposition splits revenue into two factors. This version runs that split twice — once per cohort — giving you four factors. The extra resolution is what lets you choose between a CRO experiment and a retention experiment, rather than guessing.
A buyer with at least one completed order before the period you're analyzing. Some teams add a recency cap (orders within the last 12 months) to avoid lapsed customers skewing the returning-cohort AOV downward. Pick a definition and hold it constant across comparison periods.
Shopify (or your backend OMS) for order-level truth — customer.orders_count is reliable. GA4's new_vs_returning is session-based and breaks across devices, browsers, and consent denials. Use GA4 only when you also need pre-purchase behavioral signals layered in.
At least 100 orders in the cohort for a monthly read. Below that, a few high-value or low-value baskets swing AOV by 10%+ and you'll chase noise. If you can't hit 100 monthly orders per cohort, roll the analysis up to quarters.
Compare like-for-like periods (month vs same month last year, or rolling 90 days vs prior 90 days). Always report the ratio alongside the four AOV/order numbers so anyone reading the dashboard can spot mix-shift before they over-interpret a blended swing.
Yes — heavily. Welcome discounts compress new AOV; loyalty discounts compress returning AOV. Track gross AOV (pre-discount) and net AOV (post-discount) per cohort. A growing gap between them tells you discounting is doing more of the work than the underlying basket.
New, for that period. If they reorder in the same period, the second order counts as returning. Most analytics tools default to classifying the customer once per period; the order-level classification is more precise but takes a custom query.
30-50% for most apparel and accessories, 50-80% for beauty and supplements where bundles and subscriptions stack. Under 20% means repeat buyers aren't being merchandised differently from first-timers — usually a fixable post-purchase flow problem.
Yes, and it's often more useful. Returning-cohort gross margin is typically 5-10 points higher (lower CAC amortization, better mix). Margin-decomposed cohorts surface whether revenue gains are actually profitable — important when paid acquisition is also moving.
Monthly as a standing report, weekly during a campaign window or after a major site change. The point is to catch cohort-specific moves within one decision cycle, not to refresh dashboards constantly. Pair it with a CRO and lifecycle review so the diagnosis routes to the right team.
Get an AI expert review of your site
Paste your URL — Metricuno's AI runs the same heuristic checks a senior CRO consultant would, scoring your page and prioritising the fixes that'll move conversion fastest.