Watching how visitors actually use your site

Aug 29
3 min read

Updated: 5 days ago

Introduction

Analytics reports that eighty per cent of visitors leave the service page without acting. It does not report why, so the meeting fills with theories: the copy, the price, the design, the traffic quality.

Watching five people attempt to use the page answers it in twenty minutes. They cannot find the price, they miss the button because it does not look like one, and they scroll past the form entirely. None of that appears in any report, and all of it is fixable within a day once somebody has actually seen it happen.

1. Watching how visitors actually use your site explains what analytics counts

The complementary relationship.

Numbers tell you what happened and at what scale; observation tells you why. Attempting to infer the reason from the numbers alone is where nearly every fruitless website debate begins, because any theory fits and none can be settled.

2. Start by asking five people to do something

The simplest possible version.

Give somebody unfamiliar a task — find the price of this service, book an appointment — and watch without helping. Five people will surface most of the significant problems, and the exercise costs nothing beyond an afternoon and a little awkwardness.

3. Stay silent while they struggle

The hardest part of the method.

Explaining, hinting or defending destroys the value of what you are watching. Where they hesitate is the finding itself, and the urge to help them past it is exactly what has kept the problem invisible inside the business for years.

4. Ask them to narrate their thinking

Where the useful detail comes from.

Asking people to say what they are looking for and what they expect reveals mismatches between your labels and their vocabulary. The words they reach for are frequently more valuable than the actions they take, because those words belong in your navigation and your headings.

5. Use session recordings for scale

Observation without recruiting.

Recording tools replay real visits, showing where people scroll, where they hesitate and where they abandon. Twenty recordings of a single important page will show the same pattern several times over, and repetition is what turns an observation into a finding worth acting on. Twenty recordings of your main page will show the same pattern several times, which is when it becomes a finding.

6. Use heatmaps to check assumptions

Good for one specific question.

Aggregate views of clicking and scrolling show whether people reach your content and whether they click things that are not links. They answer narrow questions well and broad ones poorly, so bring a specific question to them rather than expecting the picture to explain itself.

7. Respect privacy in whatever you record

An obligation, not an optional consideration.

Recording tools must mask personal data entered into forms, and your privacy information must disclose that sessions may be recorded. Configure this before switching anything on rather than afterwards.

8. Look for the pattern, not the anecdote

The discipline that keeps it honest.

One person doing something odd is noise; four out of five doing it is a problem. Selective attention to a single dramatic session is the main way this method gets misused.

9. Fix and watch again

Close the loop.

Make the changes, then repeat with different people. Improvements frequently reveal the next obstacle, which was hidden behind the first one and could not have been found until it was removed.

Include somebody from outside the business every time. People who work there cannot unsee the structure they already know, and their fluency with the site is precisely what makes them unable to evaluate it.

Conclusion

Use observation to explain the behaviour your analytics can only count.

Give five unfamiliar people a real task and watch in silence, ask them to narrate what they are looking for, use session recordings to see patterns at scale, use heatmaps for narrow questions about scrolling and clicking, configure privacy masking and disclosure before recording anything, act on patterns rather than single dramatic sessions, repeat the exercise after making changes, and always include somebody who does not work for the business.