Blog · Board and management reporting
A market basket analysis dashboard compares product combinations, basket value, and purchase frequency across stores and periods. This worked example shows the layout, the underlying receipts, and the checks that keep a popular pair from becoming an unsupported promotion claim.
A market basket analysis dashboard compares product combinations, basket values, and purchase frequency within a defined set of receipts. Its most useful view shows where a pair is common, whether it appears more often than expected, and how much business sits behind it. Every ratio should lead back to the same receipt population.
This is a review layout you can build from a transaction export. The broader market basket analysis method owns the full explanation of association rules; here the focus is which views help a commercial team make and check a decision.
The top of the review should show reporting dates, included stores, receipt count, and net sales. Define whether returns, canceled orders, delivery charges, and sales tax are excluded. Use the same rules for every store being compared.
Below that, use separate views for product pairs, basket value, and purchase frequency. Association-rule documentation distinguishes how often items occur together from the directional probability of buying one when another is present. A dashboard should preserve that distinction rather than label every relationship a strong bundle.
| View | Measures | Review question |
|---|---|---|
| Population | Receipts, net sales, active identified customers | Are the comparison groups consistent? |
| Product pairs | Pair count, support, confidence, lift | Does this relationship exceed ordinary product popularity? |
| Basket value | Net sales divided by receipts | Are larger baskets associated with the pair? |
| Purchase frequency | Receipts per active identified customer | Do shoppers place more orders in the same observation period? |
| Exceptions | Missing product codes, duplicate receipts, returns | Could the apparent pattern come from a file problem? |
Keep product revenue and margin detail in the product analysis dashboard, where the decision concerns the product itself rather than combinations within a basket.
A fictional convenience-store company reviews October 5 through October 11, 2026. The table contains all eligible receipts for this small illustration. Receipt value is net of discounts, excludes sales tax, and includes other products beyond coffee and bagels. Each row represents one receipt, after product lines have been grouped.
Customer labels are fictional identifiers. A Yes means the product appears at least once; buying multiple units still counts as one receipt containing the product.
| Receipt | Store | Customer | Receipt value (USD) | Coffee | Bagel |
|---|---|---|---|---|---|
| N1 | North | C1 | 10 | Yes | Yes |
| N2 | North | C1 | 7 | Yes | Yes |
| N3 | North | C2 | 9 | Yes | No |
| N4 | North | C3 | 4 | No | Yes |
| N5 | North | C2 | 8 | No | No |
| N6 | North | C4 | 6 | No | No |
| S1 | South | C5 | 11 | Yes | Yes |
| S2 | South | C5 | 5 | Yes | No |
| S3 | South | C6 | 7 | No | Yes |
| S4 | South | C6 | 3 | No | Yes |
There are 10 receipts, five containing coffee, six containing a bagel, and three containing both. Total sales are 10 + 7 + 9 + 4 + 8 + 6 + 11 + 5 + 7 + 3 = $70.
Pair support = Receipts containing both products / All eligible receipts
Coffee-to-bagel confidence = Receipts containing both / Receipts containing coffee
Lift = Coffee-to-bagel confidence / Bagel purchase rate across all receipts
These definitions follow the primary documentation for association-rule metrics. Lift equals one when the observed co-occurrence matches the rate expected under independence. Covirage is not affiliated with Oracle or mlxtend.
Round percentages to one decimal place, monetary averages to two decimal places, and lift to two decimal places. Calculate with the unrounded counts.
| Measure | North, amounts (USD) | South, amounts (USD) | All receipts, amounts (USD) |
|---|---|---|---|
| Receipts | 6 | 4 | 10 |
| Coffee receipts | 3 | 2 | 5 |
| Bagel receipts | 3 | 3 | 6 |
| Coffee-and-bagel receipts | 2 | 1 | 3 |
| Pair support | 2 / 6 = 33.3% | 1 / 4 = 25.0% | 3 / 10 = 30.0% |
| Coffee-to-bagel confidence | 2 / 3 = 66.7% | 1 / 2 = 50.0% | 3 / 5 = 60.0% |
| Lift | (2 / 3) / (3 / 6) = 1.33 | (1 / 2) / (3 / 4) = 0.67 | (3 / 5) / (6 / 10) = 1.00 |
| Net sales (USD) | 44 | 26 | 70 |
| Average basket value (USD) | 44 / 6 = 7.33 | 26 / 4 = 6.50 | 70 / 10 = 7.00 |
| Active identified customers | 4 | 2 | 6 |
| Receipts per active customer | 6 / 4 = 1.50 | 4 / 2 = 2.00 | 10 / 6 = 1.67 |
The company confidence of 60.0% looks promising until compared with the bagel purchase rate, also 6 / 10 = 60.0%. Expected pair receipts under independence are 5 × 6 / 10 = 3, exactly the observed count. The overall lift therefore gives no evidence of a stronger-than-baseline association in this file.
North and South tell different stories, but the counts are too small to justify a commercial conclusion. North has only two paired receipts. Show the count alongside the ratio and ask whether the pattern persists in a larger comparable period before proposing a promotion.
The paired receipts N1, N2, and S1 total 10 + 7 + 11 = $28, giving an average of 28 / 3 = $9.33. This describes those baskets; it does not establish that adding a bagel caused their higher value.
Suppose a product-code check finds that receipt N3 included a bagel that was classified incorrectly. Change its Bagel flag from No to Yes. Keep the receipt count, customer identifiers, and recorded dollar value unchanged.
| Measure | Original, amounts (USD) | After correcting N3, amounts (USD) |
|---|---|---|
| Coffee receipts | 5 | 5 |
| Bagel receipts | 6 | 7 |
| Paired receipts | 3 | 4 |
| Pair support | 3 / 10 = 30.0% | 4 / 10 = 40.0% |
| Coffee-to-bagel confidence | 3 / 5 = 60.0% | 4 / 5 = 80.0% |
| Lift | (3 / 5) / (6 / 10) = 1.00 | (4 / 5) / (7 / 10) = 1.14 |
| Net sales (USD) | 70 | 70 |
One corrected classification moves the association measures while leaving sales unchanged. The review needs a product-mapping exception list as well as attractive rankings.
Bring a receipt-line export with store, date, product, receipt identifier, and net sales, adding consistent customer identifiers when frequency is part of the question. Covirage maps the columns, and its tools compute and check the figures; the external AI model explains the results with every figure traced to its rows. Create a free Covirage account and bring a sample of your own data.
See revenue driver analysis for the supporting approach to understanding what moves sales.
Show the reporting period, stores, eligible receipt count, and net sales before ranking product pairs. Those totals establish the population. Then show pair counts and association measures alongside basket values, so a rare combination with an impressive ratio does not dominate the review.
No. Confidence can be high because the second product is popular across all receipts. Compare confidence with that product's overall purchase rate using lift. Then check the underlying count, product margin, and whether the relationship persists in comparable stores and periods.
Receipt data can show orders per active identified customer when it includes a consistent customer identifier. Without that identifier, report transaction counts or product purchase rates instead. Do not interpret repeat visits as retention without a defined customer population and comparable observation periods.
No. Products appearing together establish an observed relationship, not the effect of a proposed discount or placement change. Use the findings to choose a test, then measure the test against an appropriate comparison while accounting for margin, returns, and the cost of the promotion.