Calculate totals and counts using the same definition, inspect the mix and compare equivalent groups. An average summarizes data; explaining change requires opening them up.
Define what counts as an order and an amount
Before dividing, specify period, currency, order state and treatment of discounts, delivery and refunds. An average of completed orders and one of received requests answer different questions. Keep the definition beside the value so the calculation can be repeated.
NIST describes the arithmetic mean as the sum of values divided by their count. We apply that operation to a fictional example: each order has one amount from one of two categories. The figures support a calculation and do not describe an observed business.
Keep prices fixed and change quantities
In both periods, a basic order costs 100 monetary units and a special order costs 300. Each period contains ten completed orders. Only the category counts change in this example.
| Measure | Period A | Period B |
|---|---|---|
| Basic orders at 100 | 8 | 4 |
| Special orders at 300 | 2 | 6 |
| Total amount | 1,400 | 2,200 |
| Total orders | 10 | 10 |
| Average amount | 140 | 220 |
Period A totals 8 × 100 + 2 × 300 = 1,400; dividing by 10 gives 140. Period B totals 4 × 100 + 6 × 300 = 2,200; its mean is 220. The mean differs by 80 while both prices remain unchanged.
Check the weight of each group
Special orders account for two of ten in A and six of ten in B. More higher-value orders change the overall mean. In this constructed example, composition explains the difference. With real data, you would still need to inspect other changing conditions.
NIST’s weighted-mean documentation incorporates weights in the calculation. Here the weight is each group’s order count. Directly averaging 100 and 300 gives 200: that gives both categories equal weight and reproduces neither period’s distribution.
Compare using a reference mix
As an additional check, apply period A quantities to period B prices. The calculation is again 8 × 100 + 2 × 300, divided by 10: 140. Holding the mix fixed reveals that prices contributed no difference in this example. This counterfactual exercise uses chosen conditions to explain an operation.
For a real review, keep counts, total amounts and definitions by group. If a group combines differently priced products, its own mean can also move with internal composition. Start with a classification you can maintain and document its limits rather than seeking perfect segmentation immediately.
Keep the conclusion within the evidence
A supported conclusion is: “The mean rose from 140 to 220 with constant prices and a larger share of special orders.” We do not know costs, profit, unique people or reasons for choosing. Those questions require other records.
Our metrics-and-coverage guide preserves definitions. Our business-news reading guide separates data from interpretation. The same care prevents an order-mix change from becoming a claim about inflation or profitability.
Keep a record that reconstructs the calculation
- Period, currency, states and inclusion rules.
- Counts and amounts by category using the same rules.
- Overall total and mean calculated from the totals.
- Known price, discount or classification changes.
- Supported conclusion and remaining questions.
When evaluating an action such as a loyalty program, also record participation and the comparison you can support. Our loyalty-rules guide separates enrollment, activity and redemption. A higher average amount can start an investigation without automatically becoming an outcome attributed to that action.
Follow the evidence.
Sources checked October 18, 2026. Illustrative scenarios and numbers are fictional. The records and exercises are original editorial proposals, without attributed customer results or provider tests. Our editorial standards.
