/ Docs
Contents

Working with search queries

Queries is a work list, not a report. It shows what shoppers typed and where the search fell short, groups the problems by kind, and proposes one fix per row for you to confirm. At the bottom, Sentences for empty searches helps products answer the words shoppers search for and don't find.

You find it in Catalog > Search > Queries.

Choose the period

Steps

  1. In Nucleo, go to Catalog > Search > Queries.
  2. Click the period selector at the top right of the page and choose one of:
    1. Last 7 days
    2. Last 30 days
    3. Last 90 days
    4. Last 12 months (the default)
    5. A custom range: pick a start date and an end date.

The key figures and the queries table reload for the new period.

Read the key figures

The cards at the top summarize the searches made on your storefront:

CardWhat it counts
Storefront searchesSearches made in your store's search box, with the number of distinct terms.
Searches with no resultsStorefront searches that found nothing.
Searches from landing pagesVisits that started on a search results page.
Led to a clickThe share of searches that ended with a click on a result, where clicks are measured.
Orders in datasetThe orders Nucleo has read from your store, and the customers with a mapped purchase.

Below the cards, a chart per source shows the most searched terms.

Note
The storefront figures come from your store's search data imported into Nucleo and, once installed, from the storefront snippet. A line above the cards says which dates the imported terms cover. If you pick a shorter range, the imported totals still cover their whole period, and the line turns amber to say so.

Read the queries table

The Queries table lists one row per term. Use the tabs above it to show one kind of problem at a time; each tab shows how many rows it holds. To find one term, type it in the search box next to the tabs.

SignalWhat happened
Answered todayYour storefront's search found nothing, but Nucleo's search finds products or answers with a redirect.
Zero resultsNucleo's search finds nothing either.
Rarely clickedSearched at least 30 times, but fewer than 10% of the searches led to a click.
Missing stockNobody finds it, and no product mentions the word: it's goods you don't carry, not a search problem.
Ambiguous parseFew results, and the query contains a word that is a valid value of two attributes, so the search may have picked the wrong one.
Top queryOne of the most searched terms. Nothing to fix; worth watching.

Each row also shows What happened in a sentence, the number of Searches, the share that Led to a click and the Proposed fix. The Led to a click column appears only when clicks are measured for at least one term. The table shows 15 rows per page; you can switch to 30 or 50.

Fix a query

Each row proposes one fix. Clicking it takes you where the fix is made, already filled in; nothing changes until you confirm there.

Steps

  1. In Nucleo, go to Catalog > Search > Queries.
  2. Find the row. Use the signal tabs to narrow the list.
  3. Click the button in Proposed fix:
    1. Add synonym opens the synonym form with the term and its proposed target filled in. Check it and click Add synonym. See Add a synonym.
    2. Add redirect opens the redirect form in Catalog > Settings > Search > Redirects, filled in. See Add a redirect.
    3. Open in playground runs the query in the Playground so you can see what the shopper got.
    4. Send to enrichment opens Enrichment with the word. There, Nucleo tells you whether products already mention it, so it's an attribute to fill in, or whether it's goods you don't carry.

When you add a synonym or a redirect from a row, the row leaves the list.

Note
With the Viewer role, the rows that would write a synonym or a redirect show a dash instead of the button. Open in playground and Send to enrichment stay available.

Write sentences for empty searches

Some words find nothing because no product text uses them, even though the products fit, such as waterproof for a jacket whose description only lists its fabric. Sentences for empty searches proposes one short sentence for each relevant product, to add to its description. The sentence is written only when something already written about the product justifies it.

Steps

  1. In Nucleo, go to Catalog > Search > Queries and scroll to Sentences for empty searches.
  2. Select the language of the product text: IT, EN, DE, FR or ES.
  3. Click a word. Each card shows how many times it was searched with no results and how many sentences are waiting or already written.
  4. Click Propose sentences. Atomo reads 24 products of the relevant types and proposes one sentence each where a written feature justifies it. A message says how many products it read and how many it left alone.
  5. Review each sentence and its Why: line, and do one of the following:
    1. Click Approve or Reject.
    2. Click Approve all ≥ 80% to approve every pending sentence with a confidence of 80% or more, and confirm.
  6. Click Apply _n_ approved, and confirm. The sentences are written into the descriptions, and a message tells you how many products the word finds now, compared with before.

To go back to the list of words, click All words. The Pending, Approved and Applied tabs keep each group of sentences apart.

Remove an applied sentence

Steps

  1. In Sentences for empty searches, open the word.
  2. Click the Applied tab.
  3. On the sentence, click Remove.
  4. Click Remove again to confirm.

Nucleo removes exactly the sentence it added from the description, and the product is searched again without it.

Considerations for queries

  • Searches you run in the Playground are recorded too, so tests can appear in the table next to storefront terms.
  • If a message says there's no storefront term in the period, widen the range: the imported data doesn't cover the dates you picked.
  • Changing a description with sentences changes your product data, not just search. Review sentences as you would any copy.
  • You need the Owner, Admin or Editor role to propose, approve, apply or remove sentences.

Related articles