Measuring Help Center Success: The Metrics That Matter

A help center is working when two numbers move: support tickets per customer go down, and self-serve answers go up. Everything else you can measure (article views, searches, feedback ratings) exists to explain those two numbers and tell you what to fix next.

Most teams launch a help center, feel good about it, and never check whether it's doing its job. That's a shame, because the metrics are simple, most tools collect them automatically, and each one maps to a concrete action.

The north star: tickets per customer

Raw ticket count is misleading, because a growing product gets more tickets even when support is improving. Divide tickets by active customers and the trend becomes honest.

Measure it before you launch the help center, then monthly after. Any number you read elsewhere is somebody else's product and somebody else's customers, so treat your own baseline as the only one that matters. What you're looking for is a clear downward trend within a couple of months, not a specific percentage.

If tickets per customer aren't falling, the usual suspects, in order:

1. Customers can't find the help center (it's not linked where problems happen)
2. They find it but search fails them (see failed searches, below)
3. They find articles but the articles don't solve the problem

Failed searches: your content roadmap

The single most actionable metric a help center produces is the list of searches that returned nothing. Each entry is a customer telling you, in their own words, exactly what article to write next.

Check this list weekly while your help center is young. Two things to look for:

• Missing content. People search for topics you haven't covered. Write the article.
• Vocabulary mismatches. People search "receipt," your article says "invoice." Rename the article or add their word to it, as covered in the writing chapter.

HelpKit's insights surface these searches for you, and most serious help center tools do something similar. If your tool doesn't, that alone is a reason to switch.

Article-level metrics

Views tell you what matters. Your top ten articles by views are the questions your product raises most. Read them quarterly with fresh eyes, because a stale answer at the top of that list misleads more people than the bottom fifty combined.

Feedback ratings tell you what's broken. A simple "Was this helpful?" at the end of each article converts readers into reviewers. Ignore the absolute percentages and watch for outliers: an article rated far below your average either answers the wrong question or answers it badly. Rewrite it, then watch whether the rating recovers.

Zero-view articles are a naming problem. An article nobody opens usually has a title nobody searches for. Before deleting it, rename it in the customer's words and give it a month.

Signals from outside the help center

Ticket phrasing. Once the help center is live, tickets should start referencing it: "I tried the domain guide but step 3 fails." That phrasing is a healthy sign even though the ticket still exists, because the customer arrives halfway to a solution. If tickets never mention your articles, customers aren't finding them.

Search engine traffic. A public help center accumulates Google traffic on your customers' questions. Growing organic visits to articles means your help center works as marketing too, answering people who haven't bought yet.

Support team mood, honestly. The point of all this is fewer repetitive questions. Ask whoever answers your tickets whether the repetitive ones are thinning out. They'll know before the dashboard does.

A monthly 30-minute review

You don't need a reporting ritual, just a recurring half hour:

1. Tickets per customer: trending down?
2. Failed searches: write or rename the top two or three
3. Feedback outliers: rewrite the worst-rated article
4. Top ten articles: still accurate after this month's product changes?

That loop, repeated monthly, compounds. Each pass removes the biggest current weakness, and after a few months the help center answers questions you used to answer by hand every day.

Once your metrics look healthy, there's one thing left before you call the project done: making sure the launch itself was complete. Chapter 6 is the launch checklist.

Common questions

What's a good ticket deflection rate? There's no universal benchmark worth chasing, because products differ too much. Chase your own trend instead: tickets per customer falling month over month means the help center is earning its keep.

How soon after launch should metrics move? Failed searches are useful from day one. Ticket trends need a month or two of data before they mean anything. Organic search traffic builds over several months as articles get indexed and ranked.

Do I need analytics beyond what my help center tool provides? Usually not at the start. A tool with built-in insights covers the loop above. Add product analytics later if you want to connect help center visits to retention or conversion.