Blog · Finance metrics and formulas
Average orders per customer divides distinct orders by eligible customers in a stated window. Use invoice exports to remove duplicate lines and returns, compare cohorts, and prove the total.
The average order per customer formula is distinct eligible orders divided by distinct eligible customers within one stated period. It measures buying frequency, so product lines, invoice installments, and return documents must not create extra orders. Eight orders from five purchasing customers mean 1.6 orders per customer.
Average orders per customer = Distinct eligible orders in the window / Distinct eligible customers in the window
Start by deciding what eligible means. For purchasing-customer frequency, the denominator contains customers with at least one qualifying order during the window. For account coverage, it can instead contain the customer roster eligible to buy, including accounts with no orders. Both are useful, but they are different measures.
Write the definition beside the result: completed merchandise orders per purchasing customer, January 1 through March 31, 2026. That label tells a sales manager what changed when the figure moves. It also prevents an analyst from quietly switching to an opening account roster next quarter.
The numerator counts orders, not units, products, or dollars. Products per customer answers a separate cross-sell question. A large basket can contain several products while remaining a single order.
Bring customer ID, original order ID, order date, document type, status, invoice ID, and line ID. Include the cohort field you intend to compare, such as first-purchase month or customer segment. Use the order date consistently even when invoicing happens later.
These are selected rows from a US office-supply distributor's quarter. The export contains merchandise lines, an accidental duplicate, and a return document. Order O100 has different merchandise lines, so both rows are valid source records but represent the same order.
| Export row | Customer | Original order | Record |
|---|---|---|---|
| L1 | A | O100 | Merchandise line |
| L2 | A | O100 | Merchandise line |
| L3 | A | O101 | Merchandise line |
| L4 | A | O102 | Merchandise line |
| L5 | B | O103 | Merchandise line |
| L5 repeated | B | O103 | Duplicate of L5 |
| L6 | B | O104 | Merchandise line |
| L7 | C | O105 | Merchandise line |
| L8 | D | O106 | Merchandise line |
| L9 | E | O107 | Merchandise line |
| R1 | A | O100 | Return document |
For this frequency definition, the return document is excluded from order counts; it does not subtract an order. A fully canceled order would be excluded under the stated eligibility policy. Revenue treatment belongs in a separate reconciliation rather than changing an order count without explanation.
Check that each original order belongs to exactly one customer. If order IDs are reused by separate business units, use the business unit together with the order ID. Customer names alone are unreliable keys when spelling or billing addresses change.
After removing the return document and exact duplicate, collapse the retained rows to distinct customer-order pairs. The calculation follows directly from those pairs.
| Customer | Distinct eligible orders | Purchasing customer |
|---|---|---|
| A | 3 | 1 |
| B | 2 | 1 |
| C | 1 | 1 |
| D | 1 | 1 |
| E | 1 | 1 |
| Total | 8 | 5 |
| Step | Calculation | Result |
|---|---|---|
| Count distinct orders | 3 + 2 + 1 + 1 + 1 | 8 |
| Count purchasing customers | 1 + 1 + 1 + 1 + 1 | 5 |
| Divide orders by customers | 8 / 5 | 1.6 |
The result is exact. Show orders per customer to one or two decimal places in a larger report, and keep the unrounded value for subsequent calculations.
If an opening roster also contains customer F, who placed no order, roster frequency is 8 / 6 = 1.33 orders per eligible account, rounded to two decimals. The underlying orders did not change. Only the business question and denominator changed.
First create an eligible order list with one row per original order. Put customer ID in column A and order ID in column B, with the example orders occupying rows 2 through 9. The purchasing-customer version is:
=ROWS(UNIQUE(B2:B9))/ROWS(UNIQUE(A2:A9))
Microsoft's UNIQUE documentation distinguishes all distinct values from values occurring exactly once. Leave the optional exactly_once argument omitted: customers who reorder must remain in the customer count.
For a customer summary, place customer A's ID in D2 and use:
=COUNTIFS($A$2:$A$9,D2)
Fill down for the other customers. COUNTIFS counts rows meeting the supplied criteria, so this formula is valid only after the list has been reduced to one row per order. Running it directly on invoice lines overstates frequency.
Covirage is not affiliated with Microsoft.
The export bridge must close: 11 source rows minus 1 return row minus 1 duplicate row equals 9 retained merchandise lines. Those lines reduce to 8 distinct orders because O100 has multiple lines. The customer summary must also sum to 8 orders and contain 5 purchasing customers.
Keep an exception list for blank customer IDs, blank order IDs, orders with conflicting customers, and excluded statuses. Do not quietly assign all blanks to a single customer. That makes the denominator smaller and the average higher.
Compare cohorts only after checking observation time. Newly acquired accounts have had less opportunity to reorder than established accounts. Frequency also differs from the customer repeat rate, which counts customers crossing a repeat threshold rather than all their orders.
Sales operations should own the eligibility rules and customer IDs, with finance reconciling the export boundaries. Account managers can then investigate changes in ordering cadence. A rising average might reflect healthier replenishment, or customers splitting a purchase into smaller orders.
Pair frequency with cost to serve per customer before encouraging additional orders. More shipments and handling can raise service cost even when revenue stays steady. If orders arrive but fulfillment stalls, examine order-to-activation time per customer separately.
Bring an order or invoice export with customer IDs, original order IDs, dates, and document types to Covirage. Covirage maps the columns, and its tools compute and check distinct-order frequency by the cohorts supplied in your file; the external AI model explains the result with every figure traced to its rows. Create a free Covirage account and bring a sample of your own export.
See sales insights for the supporting commercial analysis.
No. Average orders per customer measures buying frequency: orders divided by customers. Average order value measures money per order: net order revenue divided by orders. A customer can place more orders while spending less on each one, so track the measures separately.
Include them when the question concerns an established eligible customer roster, such as accounts assigned to a sales team. Exclude them when measuring frequency among actual buyers. State the denominator clearly, because the two versions answer different business questions.
Only if each order always produces exactly one invoice. Split billing, partial shipments, and consolidated invoices break that assumption. Use the original order identifier where available, and document any fallback to invoice counts before comparing periods.
Choose a window that fits the normal purchase cycle and keep it consistent. Monthly reporting can work for frequent replenishment, while infrequent purchases need longer observation. Compare equally mature acquisition cohorts so newer buyers have the same chance to reorder.