Pebellslearn by building
0%
RoadmapsMatrixCoursesSandboxSign in
Content Research  ›  Lesson

Support Request Evaluation — Mining the Tickets

Support Request Evaluation 13 minClustering and counting support tickets reveals which questions repeat, and classifying each cluster as a missing, hidden or wrong page, or a product problem, tells the writer what to fix first.

Support Request Evaluation — Mining the Tickets

Joel, Nightjar's support lead, exports last month's 180 support tickets for you. Forty-six of them say some version of "I added a monitor but never got an alert." That is not 46 unlucky customers. It is one page that is missing, hidden or wrong, costing support hours every week.

The rule

Every repeated support question points to a missing, hidden or wrong page.

Support tickets are the most honest research a writer can get: real readers, stuck on real tasks, describing the problem in their own words. Evaluating them takes five moves:

  1. Collect a fixed period, such as the last month, so counts can be compared later.
  2. Cluster tickets by the question behind them, not by their wording. "No alert", "alerts broken" and "never got notified" belong together.
  3. Count and rank the clusters by volume, and note the costly ones: slow to resolve, or from customers about to leave.
  4. Classify each cluster:
    • Missing: no page answers it, so write one.
    • Hidden: a good page exists but readers do not find it, so fix its title, links or placement.
    • Wrong: a page exists but is incomplete or misleading, so fix the content.
    • Not a docs problem: a bug or a confusing product design, so pass it to the product team. Docs cannot fix everything.
  5. Measure the same cluster after your fix.

The shortcut

Count, cluster, classify. Then fix the biggest cluster first and watch its number.

Nightjar's 180 ticketsStep 1 of 4

  1. Cluster and count

    No alert received: 46. Site shown as down but loads fine: 31. Billing questions: 22. The remaining 81 spread across many small clusters.

Two traps. Writing for the loudest ticket, the angriest customer, instead of the biggest cluster. And treating every cluster as a docs problem, when some need a product fix.

Drills

Answer fast — trust the rule.

quick drill

You have a month of exported tickets. What do you do first?

quick drill

Readers keep asking how to export their reports. A good page exists, titled "Data portability". How do you classify this cluster?

quick drill

Twenty tickets report that the dashboard crashes when a monitor's name contains an emoji. How do you classify this cluster?

quick drill

How do you know your documentation fix worked?

Remember

  • Every repeated support question points to a missing, hidden or wrong page.
  • Shortcut: count, cluster, classify, then fix the biggest cluster and measure it again.
  • Trap: chasing the loudest ticket, or treating product bugs as docs problems.

Practice — level up

0/1 solved

Write your answer, then compare it with a rubric and a model answer.

ExplainTriage three clusters

Nightjar's ticket export shows three clusters: 31 tickets saying "my site shows as down but loads fine" (no page covers this); 15 asking how to add a teammate (a page exists, titled "Workspace membership"); and 9 saying the monthly invoice shows the wrong currency. Classify each cluster, say what you would do, and say how you would check that it worked.