The weekly report shows average order value up 12% on the quarter, and the growth lead has already posted a celebration in the team channel. Ten minutes later, finance replies with one line: revenue per customer is down 6%. Both numbers are correct. They are measuring different things, and the gap between them is the most useful finding in the report.
Two ratios with one numerator
AOV and ARPU start from the same revenue figure and divide it by different things.
Average order value is revenue divided by the number of orders. It describes the basket: how much a typical checkout is worth.
Average revenue per user is revenue divided by the number of users in a period. It describes the person: how much a typical customer, or active user, brings in over a month or a quarter.
The two are linked by purchase frequency. If users are defined as buying customers, ARPU equals AOV multiplied by orders per customer. A store with an AOV of 60 dollars and 1.5 orders per customer per quarter has a quarterly ARPU of 90 dollars. If users include people who never bought, a third term enters: the share of users who paid. With 20% of active users buying, the same store's ARPU across all active users falls to 18 dollars.
That decomposition is why the debate about which metric is better misses the point. ARPU contains AOV. The interesting question is what happens to the other terms when AOV moves, and a single metric cannot answer it.
Taking apart a quarter where AOV rose and ARPU fell
Here is an illustrative quarter for a DTC coffee equipment brand, built to show the mechanics.
In Q2, the store took 10,000 orders from 6,250 customers, with revenue of 520,000 dollars. AOV was 52 dollars, frequency 1.6 orders per customer, ARPU 83.20 dollars.
In Q3, the team raised the free shipping threshold from 40 to 65 dollars and added a grinder and scale bundle. Orders fell to 8,200 from 6,100 customers, and revenue came in at 477,000 dollars. AOV rose to 58.17 dollars, a gain of about 12%. Frequency dropped to 1.34, and ARPU landed at 78.20 dollars, down 6%.
The basket grew because customers stretched to reach the new threshold. Frequency fell because the same threshold killed small top up orders: the 25 dollar bag of beans that used to be a quick reorder now needed padding or got bought elsewhere. Total revenue fell 8%, even though every individual checkout looked healthier.
Without ARPU, the team would have read Q3 as a pricing win. Without AOV, they would not know why customers were worth less. Taken together, the two point at one cause and one experiment to run.
What moves each number
Levers act on different terms of the decomposition, and several push one metric up while pulling the other down.
Free shipping thresholds, bundles, volume discounts and checkout upsells raise AOV. They work on the basket. Their side effect, as the teardown showed, is that they can suppress frequency when customers consolidate purchases or abandon small orders.
Subscriptions, replenishment reminders, loyalty points and membership programmes act on frequency. Amazon Prime is the familiar large scale case: free fast delivery removes the reason to batch orders, so members buy more often in smaller baskets. That pattern tends to lower AOV while lifting revenue per member.
Acquisition mix moves both. A paid social campaign that brings in first time buyers with a 20% welcome code lowers AOV through the discount and lowers ARPU because new customers rarely order twice in the same quarter.
Price increases raise AOV almost mechanically. Whether ARPU follows depends on what happens to frequency and to conversion, which is why a price test reported on AOV alone tells half the story.
Denominators that change the answer
Most disputes between AOV and ARPU are really disputes about the denominator.
AOV has one main trap: what counts as an order. Split shipments, replacement orders, zero value sample orders and B2B wholesale orders can all sit in the same table. A brand that sends 800 free replacement orders in a quarter dilutes its AOV unless they are filtered out.
ARPU has more traps, because "user" means different things in different tools. Finance often uses buying customers, which makes ARPU equal to revenue per customer, sometimes called ARPPU, average revenue per paying user. Product analytics tools often use active users, which includes browsers who never buy. Some teams use registered accounts, which includes customers who churned two years ago.
These choices produce numbers that differ by a factor of five or more for the same store. With 30,000 active users, 6,000 buyers and 480,000 dollars of revenue, ARPU is 16 dollars per active user and 80 dollars per buyer. Both are valid. Reporting them under the same label is not.
The time window matters as much. ARPU per month, per quarter and per year are different metrics, and a product with a 60 day purchase cycle will show a jagged monthly ARPU that means very little.
Mistakes that make the comparison misleading
These errors appear again and again in DTC reporting.
- Comparing AOV to ARPU directly, as if one should be higher. ARPU per buyer is always at least equal to AOV when frequency is one or more, so the gap simply reflects frequency.
- Calculating AOV on gross revenue and ARPU on net revenue after refunds, then reading the gap as a behaviour change.
- Celebrating AOV gains from a shipping threshold without checking order count and frequency in the same period.
- Mixing new and returning customers in one ARPU during a heavy acquisition push, which drags ARPU down even when existing customers spend more.
- Letting discounts disappear from the revenue line in one report and not in another.
- Using different "user" definitions in the analytics tool and in the finance model, then arguing about which number is right.
- Reading a monthly ARPU dip as churn when it is just the purchase cycle.
Which one belongs on the dashboard
Both metrics earn a place in the decomposition, but different teams need a different headline. These questions decide which number to lead with.
- Is the decision about the checkout, such as a bundle, threshold or upsell? If yes, lead with AOV and show order count beside it. If no, go to 2.
- Is the decision about retention, subscriptions or loyalty? If yes, lead with ARPU per buyer for a period that covers at least one full purchase cycle. If no, go to 3.
- Do you need to compare cohorts or acquisition channels? If yes, use ARPU per buyer by cohort, since AOV hides how often each cohort comes back. If no, go to 4.
- Is the audience finance or investors looking at the whole business? If yes, report revenue, buyers, frequency and AOV together, so ARPU can be reconstructed and checked. If no, a single AOV trend with order count is usually enough for weekly trading.
A checklist before you report either number
Run through these before the next AOV or ARPU figure goes into a deck. Most of the items take a minute to confirm, and the one that catches the most errors is the last: if AOV times frequency does not come back to the reported ARPU within a cent or two, two of the inputs were built on different definitions, and any trend drawn from them is noise.
- Revenue definition written down: gross or net, with or without shipping, tax and discounts.
- Same revenue definition used for AOV and ARPU.
- Orders filtered for replacements, zero value orders and wholesale.
- "User" defined explicitly as buyer, active user or account, and labelled that way.
- Period long enough to cover a typical purchase cycle.
- Order count and frequency shown next to AOV whenever AOV changes by more than 5%.
- New and returning customers split when acquisition spend changed in the period.
- One illustrative decomposition, AOV times frequency, checked against reported ARPU so the numbers reconcile.
Questions teams ask about AOV and ARPU
Should shipping fees and taxes be included in AOV? Most teams exclude sales tax and VAT, because they are passed to the state, and decide once whether to include shipping charged to the customer. Shopify's analytics, for example, separates gross sales, discounts, returns, shipping and taxes, so you can rebuild AOV either way. The rule is to pick one definition and use the same one for ARPU.
Is ARPU the same as customer lifetime value? No. ARPU covers one period, such as a month or a quarter. Lifetime value adds up revenue, ideally gross profit, across the whole relationship, so it depends on retention as well. A store with a high quarterly ARPU and weak repeat rates can still have a low lifetime value.
What does Google Analytics 4 call these metrics? GA4 reports average purchase revenue, which is close to AOV, and ARPU as average revenue per active user, where active means anyone with an engaged session. Because GA4 counts non-buyers as active users, its ARPU is usually much lower than a finance team's revenue per customer.
How often should ARPU be measured for a DTC brand? Monthly ARPU is noisy for products bought every few months, such as skincare or coffee gear. Quarterly or rolling 90 day ARPU smooths purchase cycles. For subscription boxes, monthly works because the order rhythm is fixed.
Pull last quarter's revenue, order count and buyer count, and write AOV, frequency and ARPU side by side in one row. If the three do not multiply out, fix the definitions before you debate the trend.