Synthetic IndustryGet a price

Case study · Customer service

Customers kept asking the same questions. We found which ones the help centre did not answer.

The support team answered the same questions every day, though a help centre existed. We ranked what was missing, out of date or hard to find.

Who
A software company with a four-person support team and a help centre of a few hundred articles.
The problem
Customers kept writing in with the same questions, even though a help centre existed. The team could not tell which articles were missing.
What we did
Read the last three months of support tickets, grouped the questions, and matched each group to the nearest article.
The result
A ranked list of what customers ask that no article answers, each marked missing, out of date or hard to find, with example tickets.
Time
About a week.
Get a price for something like thisThe service behind it

What was going on.

The support team answered the same handful of questions every day. The company had a help centre, and a chat assistant that searched it, but customers still wrote in. Sometimes the assistant's answer was technically right and still no use to the person asking.

Nobody was counting which questions had no article, or which articles were out of date. The support lead suspected a content problem rather than a problem with the assistant, but had no way to show it.

What we found.

We worked from the last three months of tickets, which the support lead had redacted, and an export of the help centre. First we set aside everything that was never a content problem: billing questions, bug reports and feature requests. Then we grouped what was left by what the customer was really asking.

Each group got a verdict: no article, an article that was out of date, or an article that existed but was hard to find. Each group carries a count, a link to the nearest article and at least three real example tickets, so anyone can check a verdict for themselves.

The technical detail

Tickets and chats from the last 90 days were grouped by the question asked. Each group was matched to the nearest help article by content, then marked missing, out of date or hard to find. Billing issues, bugs and feature requests were kept in a separate list, with their count shown, so they did not inflate the content gaps. The group counts reconcile to the ticket export.

What we did.

How the work ranSteps in order: 1 Agree what counts: The exports, the period and what a gap is; 2 Set the noise aside: Billing, bugs and feature requests kept apart; 3 Group the questions: By what the customer was really asking; 4 Give each a verdict: Missing, out of date or hard to find; 5 Check a sample: The support lead spot-checks 20 groups; 6 Hand over the report: Ranked, with counts and the caveats. You decide at step 1 and step 5.1 Agree whatcountsThe exports, the periodand what a gap is2 Set the noiseasideBilling, bugs and featurerequests kept apart3 Group thequestionsBy what the customerwas really asking4 Give each averdictMissing, out of date orhard to find5 Check a sampleThe support leadspot-checks 20 groups6 Hand over thereportRanked, with counts andthe caveatsYou decide here How the work ranSteps in order: 1 Agree what counts: The exports, the period and what a gap is; 2 Set the noise aside: Billing, bugs and feature requests kept apart; 3 Group the questions: By what the customer was really asking; 4 Give each a verdict: Missing, out of date or hard to find; 5 Check a sample: The support lead spot-checks 20 groups; 6 Hand over the report: Ranked, with counts and the caveats. You decide at step 1 and step 5.1 Agree what countsThe exports, the period and what a gap is2 Set the noise asideBilling, bugs and feature requests kept apart3 Group the questionsBy what the customer was really asking4 Give each a verdictMissing, out of date or hard to find5 Check a sampleThe support lead spot-checks 20 groups6 Hand over the reportRanked, with counts and the caveatsYou decide here
The steps in order. A box with a coloured border is where the client decided.

How we checked it.

These had to pass before we asked for sign-off.

  • The support lead spot-checked 20 groups and had to agree with the verdict on at least 18 before sign-off.
  • Every group lists at least three example tickets, and the group counts reconcile to the ticket export.
  • Billing questions, bugs and feature requests are separated from content gaps, and the separated count is shown.

What they have now.

The support lead received the ranked report, the example tickets for each group, links to the nearest articles, the separate list of billing questions, bugs and requests, and a page on what the counts can and cannot show.

The report gave the team an order of work: the questions that cost the most tickets, and for each one whether to write, update or move an article.

What we did not do.

  • We did not write or change any article. Drafting articles is a separate job.
  • We did not change the chat assistant or the help desk.
  • We did not forecast how many tickets new articles would remove. The counts show what customers ask, not what they would stop asking.
  • We only worked with tickets the company had redacted and agreed to share.

The service behind this.

If customers keep asking things your help centre should answer, counting which questions have no answer is the first step, and it is quick.

One-off job

Help centre gap report

A ranked list of the questions customers ask that no article answers, each marked missing, outdated or hard to find, with at least three example tickets as evidence.

$230. $230 after the agreed tests are run, the results are as specified and you sign off in writing. No payment before sign-off.

Get a priceSee what is included

More case studies.

All case studies

Get a price for something like this

You are asking about: Help centre gap report.

Add detail (optional)

A web address anyone can open, with no login in it. No code or logs.

We reply by email. Sending this emails your request to our team; nothing is kept in a website database. Please do not send passwords, keys or customer records. Privacy notice.

Sending this is not an order. We check we can help, then agree what is included, safe access and the price in writing before any work. Prefer email? hello@syntheticindustry.ai.