1. Define the item and the accepted order
Name the item, variation and market covered by the test. Your working record should connect a product-relevant visit with its campaign, item, paid order and eventual contribution. A customer buying two units is different from two newly acquired customers, and a cancelled order should not remain a completed sale in the final comparison. Decide when an order is sufficiently mature for the specific question, while keeping later adjustments visible.
Separate a new-customer purchase from a repeat order when the acquisition ceiling depends on attracting someone new. Keep uncertain order status pending instead of quietly counting it as either success or failure. This gives you a consistent denominator for comparing the next source cohort.
2. Separate delivery settings from order analysis
Market and supported device settings can restrict the paid purchase. Supplier, item variation, delivered margin and order status explain what happens after it. Keep those fields distinct so a change in product mix is not mistaken for better audience selection. A low-cost item with expensive fulfilment can reduce contribution even when the source sends suitable shoppers.
Retain a compact set of product, creative, campaign and source labels through the route your store actually supports. Use order identifiers to reconcile the payment and subsequent adjustments without placing unnecessary customer details in campaign URLs. A missing label is a measurement problem to investigate, rather than evidence that the order had no value.
3. Compare one useful product question
Choose a comparison that can change the next purchase: a clearer size explanation, a different truthful product benefit or another serviceable market. Keep the underlying item and acceptance rule stable when they are not the question being tested. Record supplier or price changes because they can alter economics during the comparison.
Avoid dividing a small allowance across every item in the catalogue. Begin with a product group whose commercial result you can interpret, then expand as useful order evidence develops. A delivery check can reveal a broken route before it produces enough purchases to estimate acquisition cost; give those checks their proper, limited role.
4. Confirm what the shopper experiences
Walk through the advertised item, variation selection, cart, shipping charge, payment and confirmation. A photograph that shows several accessories can create the wrong expectation when only one component is sold. Make that distinction clear before payment. Keep delivery expectations consistent wherever the customer sees the offer, including the confirmation and support response.
Check the shopping journey on the devices the campaign is intended to reach. Include selection menus, address entry, currency presentation and any redirect to the payment provider. Record a technical failure separately from a buyer choosing not to order, then contain campaign exposure while a material fault is repaired.
5. Reconcile revenue with the cost of serving orders
Read order receipts beside the supplier and delivery costs that belong to the same cohort. Include cancellations, refunds, payment costs and support work when they affect contribution. A sales report, a media invoice and an order ledger answer different questions; establish which values each contains before combining them. Keep the comparison in one currency and on an identified period basis.
Use your store’s attribution rules consistently. WooCommerce order attribution is session-based and depends on the feature being enabled; it does not establish a complete cross-session customer journey. Reconcile what that report can show with payment and order status rather than treating an attribution label as proof of incremental profit. No store report can replace the costs needed for your own contribution calculation.
6. Check fulfilment before expanding delivery
A source that worked for a stocked item may produce a different result after a supplier substitution or shipping change. Confirm the replacement item and customer promise before adding spend. Keep the earlier buying configuration available, identify the new order period and watch whether extra demand creates dispatch or support delays.
Evaluate additional spend separately from the larger established purchase. A profitable older cohort can hide weak economics in the newest market or item variation. Compare marginal acquisition cost with a current allowance, and keep exposure manageable while refund and delivery behavior for the expansion develops.
7. Respect the store’s data and consent boundaries
Use the campaign and order information your business is permitted to process for the stated purpose. Respect applicable consent and opt-out choices, and retain only the details needed for measurement and customer service. Product and order records can explain a buying decision without exposing personal customer information in ad links or source names.
Check the actual limits of the store tools you use. A session-level attribution record should not be presented as a full customer history, and absent tracking should not automatically turn a paid order into zero value. Keep directly observed orders, estimates and unknown attribution distinguishable when deciding whether to buy more traffic.
8. Choose whether to repeat, repair or reduce the purchase
Repeat a purchase when the accepted orders and costs answer the original question within the allowance. Repair a verified item, destination or fulfilment fault when it offers a plausible improvement worth testing. Reduce exposure when the commercial loss or uncertainty is no longer affordable. The decision should reflect the item and customer promise that were actually sold, rather than a source label alone.
Keep the previous useful campaign settings, creative, product page and supplier conditions identifiable. After a material repair, verify the route and label the new period. That makes a smaller retest interpretable and gives you a practical way to reverse an expansion without discarding the evidence from the original purchase.
