Prerequisites for Reports
Select the correct store and open “Dashboard”, whose page heading is “Store overview”. Data refers to the selected period. Search must work on the storefront; visitor journeys and clicks require integration events. Dialogues and enquiries appear when those capabilities are used, and their blocks depend on plan availability.
A new store, refusal of tracking consent or missing events can leave parts of the report empty. Do not interpret “no data yet” as proof that customers did nothing. Checking the cart does not establish that completed orders are being reported.
Reading the Overview
- Select the number of days at the top and keep it consistent for comparisons.
- Review daily trends: searches, zero results and, where available, dialogues and enquiries.
- Open “Searched for and not found” using “Review searches”.
- For each frequent query, check inventory, product naming and synonyms.
- Expand “Detailed analytics” for clicks, the visitor journey and additional metrics.
| Metric | Meaning | Caution |
|---|---|---|
| Searches | Recorded requests | Not the number of unique people |
| No result | Share of searches without matches | A dash is shown when there are no searches |
| Requests | Enquiries collected during the selected period | Not only new, unprocessed enquiries |
| Chats | Conversations with the consultant | Neither sales nor a quality score |
| Average click position | The clicked result's position, starting at one | Unknown when there are no clicks |
For example, repeated “silicone” searches with no result may indicate missing inventory or the name customers use for sealant. Check real products before adding a synonym. “Returns” usually needs an informational redirect rather than a new product card.
Working With Unsuccessful Queries
You can hide a query from the working list. “Show hidden/resolved” and “Restore” bring it back. Hiding does not fix search and is not deletion of every record about that query.
If agent analysis is available, it suggests actions such as adding a product, adding a synonym or improving a card. Compare suggestions with actual inventory and today's search result. A historical unsuccessful query may already work after a catalogue update.
“Clicks on search results” includes “CTR by query” and “Most clicked products”. CTR compares clicks with impressions; a high click rate does not prove a purchase. After making changes, compare equal-length periods and consider promotions, seasonality and traffic changes.
Cart, Enquiries and Dialogues
The visitor journey shows available stages from arrival to finding a product, adding to the cart and leaving an enquiry. Without a cart event, conversion is shown as a dash. Some cart additions can originate from chat or repeat visits, so the stages do not always form a strict sequence in one session.
Open “Requests” to investigate contacts and “Chats” to examine the actual answer. Consultant ratings reflect people who submitted feedback, not every visitor's opinion. Never publish conversation contents or customer contact details in a public report.
Money: Known and Unknown
Without order reporting, search-attributed revenue is unknown, not zero. For stores reporting orders, the interface describes attribution through search interaction within a week before the purchase. An order can belong to both search and consultant groups. Revenue for individual queries must not be summed into total revenue.
Estimated lost revenue is a calculation, not money owed by customers or guaranteed future income. It needs purchase data; without that, the result is unknown. Checking search cards alone cannot justify a claim of increased sales.
Reporting Bitrix Orders
In module settings, order reporting enables transmission of the order number, amount, currency and basket contents. Check the retry queue count and last error. A selected checkbox is not confirmation that events arrived successfully. Other websites can report purchases through a supported dataLayer, AnmartoWidget.trackOrder(...) or a server integration; arrange implementation with your store developer.
A/B Checks
Open “A/B tests → Start a test”. Set a Latin-character key, name and variant traffic share, no more than half. For behaviour learning, semantic search and top-result reranking, choose leave unchanged, enable or disable; optionally set behaviour influence strength. After starting, compare “As now” and “With a change”: searches, clicks, carts and empty results. Do not declare a winner after the first few visitors or when the report says data is insufficient. Stopping ends traffic splitting but preserves collected figures.
When Figures Do Not Match
Check the store, period, tracking consent and transmission of the relevant events. Perform one reference search and click, but do not place a fictitious order to make a chart look complete. Provide support with the action time and expected metric, without personal data. Continue with search rules, cart, limits and security or all articles.