Measurement Planning
Basket and merchandising analytics
Use product pairings, recommendation results, category routes and onsite search evidence to guide merchandising decisions.
Basket and merchandising analytics shows which products to display together and how shoppers find them. Completed orders suggest possible pairings, storefront interactions reveal discovery problems, and a suitable comparison tests whether a change helped.
Products bought together in the past do not prove that displaying them together will create extra sales.
Match the evidence to the decision
| Decision | Evidence to examine | Limit |
|---|---|---|
| Offer a companion product | Qualifying orders containing both items | Co-purchase does not establish incremental demand. |
| Change a recommendation | Actual displays, clicks and commercial outcomes | Clicks alone do not measure the effect of the placement. |
| Change category navigation | Comparable entry routes and relevant products reached | Different routes may attract shoppers with different intentions. |
| Improve onsite search | Empty searches, searches with no clicks and inspected results | A result without a click is not necessarily unsuitable. |
Keep the counting unit beside every measure. An order can hold several items; a session can hold several searches or page views. Never combine orders, sessions and events in one rate.
Find pairings in orders
Count each qualifying order once for every product pair it contains, regardless of units bought. Compare a pair's order count with orders containing each product separately. A popular item appears in many pairs simply because it sells often.
Inspect whether the items are useful complements, alternatives or parts of a larger purchase, and check availability before proposing an offer.
Shopify lists an Items bought together report among its order reports. A store on another platform can build an order-level pairing view from its own records under a stated order-status rule.
Assess recommendations separately
Record who was eligible to see a recommendation, what was actually displayed, clicks and subsequent qualifying orders. Shopify's Search & Discovery reports give click and purchase rates for product recommendations; treat these as engagement evidence.
To estimate an effect, compare the commercial outcome for groups assigned to different experiences before exposure, where feasible.
Check what each report covers
Shopify’s Behavior reports include dedicated views for search queries, searches with no clicks, searches with no results, search conversions over time, product recommendation conversions over time and recommendations with low engagement. Together these views show whether a discovery issue is concentrated in a query, a recommendation surface or a change over time; they do not explain the cause on their own.
On desktop, open Analytics > Reports, select the Category filter and choose Behavior. In the Shopify app, open Analytics > Reports, tap the filter icon, choose Category, then select Behavior.
The Search & Discovery app reports give click rate and purchase rate for both search and recommendations. For search they also show searches by query, searches with no results and searches with no clicks. For recommendations, a low-engagement report focuses on customer engagement with recommendations on the store’s top-selling products.
The metrics shown in Search & Discovery cover the last 30 days. If a decision needs another date range, Shopify directs users to the full performance reports under Analytics > Reports.
Use the report labels to keep the outcome tied to the discovery surface: search purchase rate concerns products discovered through search results, while recommendation purchase rate concerns products discovered through recommendations. These are distinct measures, so a change in one should not be treated as evidence that another surface improved.
Key metrics from Shopify’s Search & Discovery reports
- Searches with no results
- Reported separately
- Searches with no clicks
- Reported separately
Inspect discovery routes
For category navigation, define the entry point, relevant product destination and observation window. Use an analytics view that supports the route question, and check its scope before interpreting the results. Observed routes alone do not reveal why a shopper chose one.
For search, inspect queries with no results and queries whose results received no clicks. Shopify reports these groups separately. Its Search & Discovery reports cover results-page activity, excluding predictive-search interactions.
Check the actual results, product visibility and stock before changing product information or search settings.
Record each proposed change, its intended outcome, the comparison used and concurrent changes such as promotions or stock gaps. Recheck the same outcome after launch and describe the result as observational unless the comparison supports a stronger conclusion.
In this guide
- Analysing products commonly bought togetherCount product pairs in qualifying orders, compare directional rates and shortlist useful combinations for a measured trial.
- Measuring the effect of a product recommendationSeparate recommendation exposure and clicks from sales effect, then compare eligible groups on a defined commercial outcome.
- Comparing category navigation pathsCompare category routes from a defined entry task to relevant products, with clear counting rules and path-analysis limits.
- Identifying search terms that return no useful productsFind onsite searches with empty or weak results, inspect the products shown and choose a measured fix.
