Measure and optimize search
Search optimization works best as a loop: observe, diagnose, change one thing, verify, and measure. Avoid changing the catalog profile, theme, controls, and rules together because the result becomes impossible to explain.
Analytics identifies the problem worth investigating.
Preview helps explain the result before you change live behavior.
Set an operating cadence
Section titled “Set an operating cadence”Daily during launch or a major campaign
Section titled “Daily during launch or a major campaign”- Confirm search is live.
- Check Sync and Index for a failed or partial job.
- Check the Help page for degraded systems.
- Run the campaign’s highest-value queries.
- Confirm campaign rules and redirects are active.
- Check mobile quick search and the full page.
Weekly during normal operation
Section titled “Weekly during normal operation”- Review Home.
- Check monthly sessions and upcoming traffic.
- Open Analytics for the last 30 days.
- Review Search insights.
- Select one high-value problem.
- Reproduce it in Preview.
- Make one targeted change.
- Test the storefront.
- Record the date, reason, and expected metric.
Monthly
Section titled “Monthly”- Compare 30- and 90-day trends.
- Review search profile and catalog health.
- Remove obsolete rules and redirects.
- Review control engagement and filter use.
- Review product opportunities.
- Confirm plan usage and limits.
- Review theme and translation changes.
- Decide whether the store is ready for an A/B test.
Start with a question
Section titled “Start with a question”Do not browse charts without a decision in mind. Good operating questions include:
- Which high-volume queries return no products?
- Which important queries have weak click-through?
- Are shoppers finding relevant products near the top?
- Which discovery surface produces carts or revenue?
- Which filters and controls are actually used?
- Which products receive exposure but weak outcomes?
- Which products perform well despite low exposure?
- Did a known change improve the intended behavior?
Read Overview as a health check
Section titled “Read Overview as a health check”Total searches
Section titled “Total searches”Use it to understand demand and traffic. A change can reflect seasonality, marketing, site traffic, or search adoption. It does not measure relevance by itself.
Click-through rate
Section titled “Click-through rate”CTR is the share of searches that led to a product click.
Investigate a decline with:
- Average click position.
- Time-to-click.
- No-result rate.
- Query mix.
- Product-card presentation.
- Filter and control behavior.
Conversion rate
Section titled “Conversion rate”Conversion rate connects search with attributed paid orders. It can change because of search quality, price, inventory, checkout, traffic quality, promotions, or attribution timing.
Revenue from search
Section titled “Revenue from search”Use Attribution to separate direct VIBE, cart-add-confirmed, click/view, and control revenue. Compare the definition and time window before comparing with another analytics tool.
No-result rate
Section titled “No-result rate”Lower is normally better, but a small number of honest no-result searches can be healthier than returning irrelevant products for everything.
Use the query table to distinguish:
- Product not indexed.
- Product not stocked.
- Shopper language gap.
- Typo or synonym need.
- Collection or variant capability gap.
- Query that should redirect to content.
Use Engagement to locate friction
Section titled “Use Engagement to locate friction”Viewed products
Section titled “Viewed products”Shows the breadth of products reached through discovery. A rise can mean shoppers are exploring more, but also that they need more clicks to find the right item. Read it with time-to-view and conversion.
Searches per session
Section titled “Searches per session”A moderate repeat rate can indicate healthy refinement. A sharp increase with weak clicks can indicate shoppers are repeatedly reformulating unsuccessful queries.
Unique products per search and session
Section titled “Unique products per search and session”These show discovery breadth. Use them after adding Explore, collections, filters, or visual discovery.
Average time to view
Section titled “Average time to view”Lower is normally better. A rise can indicate slower responses, unclear cards, low relevance, or more complex browsing.
Product discovery by area
Section titled “Product discovery by area”Compare Quick search, Search page, Explore, Similar, Taste, and other reported surfaces where present.
For each area review:
- Requests and share.
- Cart adds and revenue.
- Sessions.
- Average response time.
- Change against the previous period.
Do not disable a low-volume surface until you understand whether it serves a smaller but high-value use case.
Average click position
Section titled “Average click position”Lower means shoppers usually click nearer the top. Track it after profile, catalog, or ranking changes.
Average time-to-click
Section titled “Average time-to-click”Lower suggests shoppers recognize the useful result faster. A theme-card redesign can change this even when ranking stays the same.
Use Attribution carefully
Section titled “Use Attribution carefully”The conversion funnel is:
- Searches.
- Results clicked.
- Added to cart.
- Purchases.
Locate the largest unexpected drop.
| Drop | Investigate |
|---|---|
| Search to click | Relevance, no results, cards, response time |
| Click to cart | Product fit, price, availability, product page, selected variant |
| Cart to purchase | Checkout and commercial factors beyond search |
Paid-order attribution arrives after search events. Avoid judging the most recent day before orders and workers settle.
Use Search insights to choose work
Section titled “Use Search insights to choose work”Top search queries
Section titled “Top search queries”Prioritize terms with both high volume and commercial importance. A low-volume query can still matter when it represents a high-value product or critical customer need.
Query-product winners
Section titled “Query-product winners”These are products with strong query-specific response. Use them as evidence for natural ranking or a behavioral rule, not as a reason to pin every winning product.
Products needing attention
Section titled “Products needing attention”These products receive impressions but weaker clicks, orders, or revenue efficiency.
Check:
- Is the product relevant to the queries showing it?
- Is the card image or title clear?
- Is the price or availability competitive?
- Is a rule overexposing it?
Hidden gems
Section titled “Hidden gems”These products perform well despite lower exposure. Consider improving source data, adding a scoped boost, featuring the collection, or using them in a campaign.
Revenue per impression
Section titled “Revenue per impression”This measures attributed revenue efficiency relative to exposure. Use it with absolute volume so one high-value order does not dominate a tiny sample.
Top filters used
Section titled “Top filters used”High use shows shopper demand for exact refinement. Confirm the filter values are clean and that mobile filter access is easy.
Improvement decision tree
Section titled “Improvement decision tree”Product absent
Section titled “Product absent”Fix eligibility and index.
Product present but broadly misunderstood
Section titled “Product present but broadly misunderstood”Improve Shopify data or the shared search profile.
Equivalent shopper word is missing
Section titled “Equivalent shopper word is missing”Add a synonym.
One campaign needs business priority
Section titled “One campaign needs business priority”Use a scoped boost, bury, pin, or redirect.
Preference-based browsing is difficult
Section titled “Preference-based browsing is difficult”Improve semantic controls and starting collections.
Factual narrowing is difficult
Section titled “Factual narrowing is difficult”Improve Shopify filters and structured values.
Cards do not communicate the result
Section titled “Cards do not communicate the result”Improve theme or VIBE presentation without changing ranking.
Change log template
Section titled “Change log template”For every material optimization, record:
- Date and owner.
- Store and theme.
- Exact problem.
- Evidence and date range.
- Affected query, product, surface, or audience.
- Setting changed.
- Why this layer was selected.
- Preview result.
- Storefront QA result.
- Expected metric and review date.
- Rollback.
This prevents old campaign rules and unexplained profile changes from accumulating.
When to use an A/B test
Section titled “When to use an A/B test”Run a test when:
- The index is healthy.
- Installation is stable.
- Major rules and controls are stable.
- Store traffic is sufficient.
- You want to compare VIBE against Shopify native search.
Do not use an A/B test to diagnose a broken installation or obviously missing catalog data.
During the test:
- Avoid publishing a new theme.
- Avoid broad catalog-profile changes.
- Avoid major merchandising campaigns unless both groups are affected fairly.
- Review both groups, not only one headline metric.
- Let the configured completion rule work unless shopper harm requires stopping.
After the test, record the test period, traffic context, major store events, and full metric table.
Avoid false conclusions
Section titled “Avoid false conclusions”- Correlation is not proof of causation.
- A metric can move because traffic mix changed.
- A small sample can produce large percentages.
- Revenue can lag searches.
- Averages can hide a few high-volume problem queries.
- A rule can improve one query while harming another.
- A theme change can affect click behavior without changing relevance.
- An inventory event can affect conversion without search changing.
Use exact query and product evidence alongside aggregate metrics.