Guide · Updated
How to find what's missing from your help center
A docs gap is a question customers ask that your help center doesn't answer, answers badly, or answers in words nobody searches for. Your support inbox already holds the list. This guide shows where gaps show up, how to run a gap analysis in an afternoon, and how to keep finding them without a docs team.
What is a docs gap?
A docs gap is any customer question your help center can't answer well. There are three kinds: missing (no article exists), unclear (an article exists but customers still ask), and outdated (the article describes how the product used to work). All three end the same way, with a ticket you answer by hand.
A fourth kind hides inside the first two: vocabulary. Customers ask about “late fees” while your article is titled “Overdue invoice automation”. The answer exists, but nobody can find it, so it behaves like a gap.
Where do docs gaps show up?
Gaps show up wherever customers ask questions in their own words. The strongest signal is support conversations, because each one cost someone time. Search and assistant logs add volume, and feedback votes point at specific articles.
| Source | What it tells you | How to collect it |
|---|---|---|
| Support tickets and chats | What people couldn't solve alone | Export the last two or three months |
| Assistant questions it couldn't answer | Missing articles, in customer words | Your AI assistant's unanswered list |
| Help center searches with no results | Missing articles and missing vocabulary | Help center or site search analytics |
| “Not helpful” votes | Articles that exist but fail | Article feedback, ideally with a reason |
| Onboarding and sales calls | Questions before people even sign up | Call notes, recordings, your own memory |
| Community or forum threads | Questions customers answer for each other | Your community or public forums |
Start with tickets. If you only have time for one source, it should be the one where every question already cost you money.
How do you run a docs gap analysis in an afternoon?
Export recent support conversations, tag each one by the article an agent would send, and count. Questions with no article to send are your missing docs. Questions with an article that didn't help are your unclear docs. Two or three months of tickets is usually enough to see the pattern.
- Export two or three months of tickets and chats into a spreadsheet, one row per conversation.
- Add a column: “Which article would I send?” Write the article title, or “none”.
- Add a second column: “Did that article answer it?” Yes, partly, or no.
- Group the rows by article. Count how often each one comes up.
- Rows marked “none” become new articles. Rows marked “partly” or “no” become rewrites.
- Sort both lists by count. The top five are your next week of writing.
This takes a few hours the first time. It is the highest value docs work most small teams never do.
How do you group questions into articles?
Group questions by the page you would send, not by the exact wording. Customers phrase the same need ten ways; one good article answers all ten. Writing one article per raw question produces a help center full of near duplicates that compete with each other in search.
Here is how it looks for Ledgerloop, the invoicing SaaS we use in examples:
- “Can I charge a late fee?”, “add penalty to unpaid invoice”, “automatic overdue fee” → Add late fees to overdue invoices
- “invoice every month automatically”, “recurring billing for a client” → Set up recurring invoices
- “download invoice as PDF”, “print invoice”, “send invoice without email” → Export invoices as PDF
- “client paid half”, “record partial payment” → Record partial payments
Use the customers' words in titles and headings. That one change closes many vocabulary gaps without writing anything new.
Which docs gaps should you fix first?
Fix the gaps that cost the most first: high volume questions, questions that block setup or payment, and questions about features you just shipped. A short rewrite of a heavily used setup guide often beats a new article on a rare edge case.
| Gap | Volume | Cost when unanswered | Priority |
|---|---|---|---|
| Setup step customers get stuck on | High | Lost activation | First |
| Billing or plan limit question | Medium | Refunds, churn, angry tickets | First |
| Feature shipped last week | Rising | Every new user asks | First |
| Integration edge case | Low | One long ticket each time | Later |
| Rare advanced setting | Low | Small | When convenient |
If two gaps tie, pick the one where a wrong guess by the customer would do the most damage.
How do you write the article so the gap actually closes?
Write the title in the customer's words, answer the question in the first two sentences, then give the steps. Afterwards, test it: search your help center and ask your assistant the original question, phrased the way the customer phrased it. If it doesn't come up first, the gap is still open.
- Title: the question or task as customers say it
- First lines: the direct answer, including “no” when the product can't do it
- Then: steps, limits, and what to do if it doesn't work
- Link: the related articles a customer will need next
“No, Ledgerloop doesn't support that yet” is a valid article. It closes the gap, stops the guessing, and gives your assistant something honest to cite.
How do you keep finding gaps after the first pass?
Make gap finding a weekly habit instead of a yearly project. Spend twenty minutes reading the questions your assistant couldn't answer, the searches with no results, and any “not helpful” votes from the past week. Add the new themes to your list and write the top one.
Also check after every release. New features create new questions within days, and changed features make old articles wrong. Gaps from customers and drift from code are two halves of the same job: keeping the help center true to the product.
How does usedocs find and fill gaps?
usedocs groups the questions its assistant couldn't answer, “not helpful” votes, and tickets imported from Zendesk, Intercom, Help Scout, Crisp, Plain, HubSpot, or a CSV into ranked gaps. Each gap gets a drafted article written from your existing docs and sources, and it says what it couldn't confirm instead of making it up.
Drafts wait in one review queue next to edits proposed from merged pull requests, so missing docs and outdated docs get fixed in the same place. To see your own gaps first, run the free docs scan at Documentation gap analysis on your help center.
FAQ
How often should I review docs gaps?
Weekly for an active product, and right after launches, pricing changes, or big releases. Twenty minutes a week is enough on a small team.
What's the difference between a gap and outdated docs?
A gap is a question with no good answer. Outdated docs answer confidently but wrongly, because the product changed. Both cause tickets; outdated docs also cause wrong actions.
Can AI find docs gaps for me?
AI is good at clustering real questions and drafting articles from them. It shouldn't invent the list: real customer questions are the signal.
What's the fastest way to close a gap?
Add a short direct answer under the right heading in an existing article, using the customer's words, then link to deeper detail.
How do I know a gap is closed?
Search your help center and ask your assistant the original question. Then watch whether that question stops showing up in tickets over the next few weeks.
Use usedocs for this
usedocs groups unanswered questions, “not helpful” votes, and imported tickets into ranked gaps and drafts an article for each, next to edits proposed from your merged pull requests.