Mapping Lay Search Terms to Four Help-Center Articles
Fernway Fitness runs a subscription app. Its help center has four articles: “End your plan,” “Pause your plan,” “Change your plan,” and “Update payment details.” Search logs for the last 30 days show 1,240 searches for stop my membership, 610 for freeze, 480 for hold, 300 for downgrade, and 220 for cancel. Click-through is 38% for “End your plan,” 11% for “Pause your plan,” and 6% for “Change your plan.” The taxonomy specialist has 10 retagging slots this sprint and cannot change article titles because legal review is closed. Legal also requires that the word cancel be reserved for statutory cancellation rights and not used as a UI synonym for ending a plan. Product uses membership in the app but subscription in billing emails. A Spanish-language reviewer notes that pausa and congelar are used interchangeably by users, though the product treats them as one action. The specialist must decide which articles get which search synonyms, and whether downgrade should point to change or end.
1. Assign the five queries to articles, justifying each by the outcome the user wants rather than by surface similarity.
Read the worked answer
The mapping should follow outcomes. Stop my membership belongs to “End your plan”: the user wants the service to stop, which is the end state. Freeze and hold belong to “Pause your plan”: both imply a temporary stop with the account retained, and the product treats them as one action. Downgrade belongs to “Change your plan,” not “End your plan,” because a downgrade keeps the account open and changes the tier; billing continues at a lower level. Cancel is the difficult one. The user intent is usually to end the plan, but legal has reserved cancel for statutory cancellation rights and barred it as a UI synonym. The safe solution is a search redirect from cancel to “End your plan” without displaying cancel as an approved synonym on the article; the redirect serves the user without rewriting the UI.
The Spanish note does not change the English mapping. Pausa and congelar both point to the same pause action, so they can share one article and one concept ID.
2. Why is click-through alone a weak signal here? What additional evidence would you request before finalizing the tags?
Read the worked answer
Click-through measures what happened after a result was shown, not whether the result was correct or whether the user’s problem was solved. A low 11% click-through on “Pause your plan” may mean users do not recognize the word pause as their freeze or hold; it may also mean the article appeared below a more prominent but wrong result. The 38% figure for “End your plan” could reflect position bias, title match, or genuine intent, and it says nothing about whether users later reopened a ticket or searched again.
Before finalizing, I would request the search-result impressions and average position for each query, the zero-result rate, refinement queries within the same session, and the rate of repeat searches within 24 hours. I would also want the support-ticket reason codes for the same period, because a user who searches freeze and then opens a billing ticket may have been routed to the wrong article. Those signals distinguish a vocabulary mismatch from a ranking problem.
3. With 10 retagging slots, propose an allocation and explain what you would leave out.
Read the worked answer
I would allocate by outcome clusters, not by raw query volume alone. A defensible split is: four slots for “End your plan” (stop, end, quit, close), three for “Pause your plan” (freeze, hold, suspend), two for “Change your plan” (downgrade, switch), and one reserve slot for a term that emerges in refinement data. Cancel is handled by redirect, not by an approved displayed synonym, so it does not consume a slot. Membership and subscription are handled as a cross-article note or a global synonym rule if the platform supports it, because both terms can appear with end, pause, and change queries.
I would leave out generic terms such as account and billing because they are too broad and would reduce precision across all four articles. I would also leave out the Spanish pair pausa and congelar from the English slot count; they belong in the Spanish synonym set, where they can share one concept. The reserve slot matters because 10 slots is a hard constraint and the logs show a long tail that may contain a higher-value mismatch than a fourth synonym for end.