Guide · Updated · 5 min
What “Was this helpful?” votes tell you
The little “Was this helpful?” prompt at the bottom of a help article is easy to add and easy to ignore. Read properly, it's the cheapest signal you have about which articles fail customers and why. Here is what the votes mean, how to get more from a “no”, and how to turn them into fixes.
Short answer
A “no” vote usually means the article is wrong, unclear, about the wrong question, or describing something the product can't do. Ask one short follow up (“What were you looking for?”) after every no, read those answers weekly, and treat articles with rising no votes after a release as likely outdated. Raw percentages matter less than changes and the words people type.
What does a “no” vote usually mean?
A “no” means the reader didn't get what they came for, which can happen for very different reasons. Before fixing anything, work out which one it is, because each has a different fix.
| What happened | What the fix is |
|---|---|
| The article is outdated | Update it to match the product |
| The article is unclear or skips a step | Rewrite or add the missing step |
| The reader wanted a different question | Write the article they wanted, link it here |
| The product can't do what they wanted | Say so plainly in the article |
| They wanted a person, not an article | Make contact options easier to find |
The third and fourth are the most common and the easiest to miss: the article is fine, it just isn't the answer this reader needed.
Why should you ask a follow up question?
Because a bare “no” tells you something failed but not what. One short question after the vote, “What were you looking for?”, turns the vote into a sentence in the customer's own words. Those sentences are some of the best docs input you'll ever get: specific, unprompted, and written at the exact moment the docs failed.
Keep it optional and keep it to one field. A form with three questions gets skipped; a single text box with a clear prompt gets answered more often than you'd expect.
How should you read the numbers?
Watch changes and patterns, not the overall percentage. A helpfulness score depends on the topic: billing articles collect more “no” votes because people arrive frustrated, and troubleshooting articles collect them when the fix didn't work. Comparing a billing article with a feature overview tells you little.
- A jump in “no” votes right after a release usually means the release made the article wrong.
- Many views, few votes, and many follow up tickets means people don't bother voting; check the tickets.
- Low votes on a rarely viewed article aren't statistically meaningful; read the comments instead.
- The same follow up answer repeated across articles points at a missing article.
Which articles should you look at first?
Start with articles that combine high traffic and a rising share of “no” votes, then any article whose votes changed right after a release. After that, read every follow up answer, whatever the article, because a single clear sentence can reveal a missing article that no metric would.
- High traffic, rising “no” share.
- Any article whose votes changed after a release.
- Follow up answers that mention something the docs don't cover.
- Billing and setup articles with any “no” votes at all.
How do you turn votes into fixes?
Read the week's “no” votes and follow up answers in one sitting, sort each into outdated, unclear, wrong question, or product limit, and fix the top few. Twenty minutes a week keeps the list short. Letting votes pile up for a quarter turns them into a report nobody reads.
For Ledgerloop, the invoicing SaaS we use in examples, a week might show three “no” votes on the recurring invoices article with follow ups like “how do I stop a recurring invoice?”. The article explains starting a schedule but not stopping one. One new section fixes all three.
Should you show helpfulness scores to readers?
Usually not. A visible score invites voting on the score rather than the article, and a low number on an important page makes the whole help center look worse than it is. Keep the prompt visible and the numbers internal.
What you can show readers is that feedback matters: a short “Thanks, we'll use this to improve the article” after a vote, and a last updated date that changes when you act on it.
How do votes fit with other docs signals?
Votes tell you which existing article failed. They don't tell you about questions with no article at all, because nobody can vote on a page that doesn't exist. Pair them with failed searches, tickets, and the questions your assistant couldn't answer to see both halves: articles that exist but fail, and articles that should exist but don't. How to find what's missing from your help center covers the second half.
Votes also lag behind releases. By the time “no” votes rise, customers have already been misled. Checking articles against each release, as in How to find outdated help articles, catches the same problem earlier.
How does usedocs use helpful votes?
usedocs help center articles can end with “Was this helpful?” (a switch in Customize). A “no” asks what the reader was looking for, and that sentence becomes a gap, grouped with similar questions from the assistant and imported tickets and ranked by how often it comes up. Each gap gets a drafted article in the review queue. Reporting shows views and helpful percentage per article, so you can see which pages need work first.
FAQ
Is a “Was this helpful?” widget worth adding?
Yes, if someone reads the results weekly and the “no” vote asks one follow up question. Without both, it's decoration.
What's a good helpfulness score?
There's no universal benchmark; it depends heavily on the topic. Track changes per article over time rather than chasing a number.
Why do billing articles get more “no” votes?
People often arrive frustrated or hoping for an exception the article can't give. Read the follow up answers to tell real problems from disappointment.
Should “no” voters be offered a way to contact support?
Yes. After the follow up question, show the contact option. A reader who just said the article failed is the one most likely to need a person.
How many votes do I need before acting?
For a single clear follow up comment, one. For conclusions from percentages, wait for enough votes that a couple of people don't swing the number.