Time on Page and Why It Lies

The number that looks like a verdict

Open the pages report in Analytics, and beside each page sits a figure for average time on page. It looks like a measure of how interesting the page is.

Owners read it that way. A long time means a good page; a few seconds means people glanced and left.

On many small business sites, the page with the phone number on it shows one of the lowest figures of all.

That is not because the page is failing. It is because of how the number is calculated.

How Analytics times a page

Universal Analytics records the moment each page loads. It has no way of knowing when somebody stops reading.

So it works out time on a page as the gap between that page loading and the next page loading in the same visit.

The last page of a visit has no next page. Its time cannot be calculated, so Analytics leaves it out.

The average is the total timed seconds divided by the views that were not exits. The last page of every visit simply does not count.

Who disappears from the figure

Leaving out the last page sounds like a technicality. In practice it removes some of the most valuable visitors a site has.

• People who found the phone number and rang.
• People who read everything on one page and left satisfied.
• People who arrived from search, got their answer and closed the tab.
• People who sent a form that shows its thanks on the same page.
• People who copied your email address into their own email program.

The first is the group that matters most to a local business. Someone who lands on the contact page, reads the hours, taps the number and calls has used the page perfectly, and Analytics records no time at all.

Why a page can show zero

When every view of a page is the last of its visit, Analytics has nothing to time, and it shows the average as zero.

When most views are exits, the average comes only from the minority who went on to another page.

Those are often people who had not yet found what they needed and kept looking. So the figure can describe the least satisfied visitors best.

The page that works best, by answering the question at once, can look like the worst on the site.

Session duration has the same flaw

Average session duration, shown on the overview, is calculated the same way, from the first recorded moment of a visit to the last.

A visit that views a single page and records nothing else has a duration of zero, however long the person spent reading.

That is why a site where people find what they need on the first page, as a good local business site should, can show short visits and a high bounce rate at the same time.

A worked example

A well drilling company with two rigs in the Shuswap was about to redesign its service area page, because Analytics showed visitors spending only a few seconds on it.

The page listed the towns and rural areas the company drilled in, with the phone number at the top. Most of its visitors came from local searches.

A closer look showed that almost every view of the page was also the last page of its visit. The few seconds came from the handful who clicked on elsewhere.

So its web company set up events, recording each tap on the phone number or the email address, and a goal for the quote form's thank-you page.

Within a few weeks, the service area page turned out to be one of the places on the site where people most often tapped to call. The redesign was cancelled.

Better signs that a page works

The question is not how long people stayed. It is whether they did what the page is for.

For most business pages, that means a call, a form or a visit, and each can be measured more directly than time.

Taps on a phone number can be recorded as events, through Google Tag Manager or a little code added by whoever maintains the site.

Form submissions can be counted as goals, most reliably by sending people on to a separate thank-you page. Clicks on directions, an email address or a booking button can be recorded the same way.

Scroll depth for long pages

For long pages, such as guides or detailed service pages, it helps to know how far people read.

Tag Manager can record when a visitor scrolls past set points on a page, such as a quarter, half and three quarters of the way down.

That shows whether people reached the part that matters, which time on page cannot tell you.

Keep it modest. A few markers on the pages that matter are more useful than measuring everything.

Ask the callers

Not everything worth knowing shows up in Analytics, and the simplest check costs nothing.

Ask new callers how they found you and what they were looking at before they rang. Write the answers down for a month.

A page that keeps coming up in those answers is doing its job, whatever its average time says.

The counter-case

Time on page is not useless. For pages that are rarely the last of a visit, such as an overview that leads on to detail, it is a fair rough guide.

Comparing a page with itself over time can also be informative, since the distortion is similar from month to month.

And a page where people spend a few seconds before moving on may genuinely be failing to say what it should. That time is real, because a next page was loaded.

But never judge a page people leave from by how long they seemed to stay. Judge it by what they did.

The check

1. Find the pages with the lowest time on page.
2. Check the exit rate beside each one.
3. Record phone taps as events.
4. Count form submissions as goals.
5. Add scroll markers to long pages.
6. Ask new callers what they were reading.
7. Judge pages by actions, not minutes.

Step three changes the picture fastest for a local business. Phone taps happen on exactly the pages Analytics cannot time, so they fill the gap where the figure was blind.

Checking that the tracking itself still works is covered in opening Analytics for the first time in a year.

Frequently asked questions

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.