Document analytics explained: what time on each page tells you

Time on each page tells you where a reader's attention went: which pages they lingered on, which they skimmed and where they stopped. It does not tell you whether they agreed, understood or plan to act, and on its own a single number can mislead. This guide explains what document analytics measure, how reading time is counted, how long a page should take, the traps (idle tabs, forwarded links, skimming, downloads) and how to turn the numbers into a sensible next step.

What document analytics measure

When you share a document as a link rather than an attachment, the page that shows it can record what happens there. Tools differ, but the core is similar. Using Breathier as the example, each visit records:

• Who, if the link asks for an email address (the reader types it; it is not verified).
• When the visit happened.
• The device: phone, tablet or computer.
• The country the visit came from.
• Total reading time for the visit.
• Time on each page: each page of a PDF or Word file, each slide of a presentation, each sheet of a spreadsheet.
• Whether the file was downloaded.

On top of individual visits, a chart shows the average time on each page across all visits, and you can filter visits by link. Your own look at your link is not counted, so checking that it works does not muddy the figures.

No tool can see the reader's face, so everything below is about inferring attention from behaviour, carefully.

How reading time is counted

The most important thing to know about any time-on-page figure is what stops the clock. If a tool simply measured the gap between opening and closing a tab, a deck left open over lunch would look like the most gripping document ever written.

Browsers help here. The Page Visibility API lets a web page know when it is hidden, which MDN describes as the tab being in the background, the window minimised or the device's screen off. Web analytics use the same idea: Google Analytics defines user engagement as time a page is in focus or an app is in the foreground.

Breathier counts reading time only while the tab is visible and the reader has scrolled, tapped or moved in the last five minutes. That rule has two consequences worth remembering:

• A tab left open can add up to five minutes of time after the last movement before the clock stops. Someone who reads page 4, then answers the phone, may show a long page 4.
• A very slow reader may be undercounted. Someone who studies one dense page without touching anything for more than five minutes stops accruing time until they move again.

How long should a page take?

To judge whether someone read or skimmed, you need a baseline. A good one is a 2019 review of 190 studies by the psychologist Marc Brysbaert, published in the Journal of Memory and Language. It estimates that adults reading English non-fiction silently average 238 words per minute, with most falling between 175 and 300. It also notes that people reading in a second language, and older adults, tend to read more slowly.

Here is what that means for typical pages:

Charts, tables and financial statements don't follow word counts; people study them, so expect more time than the words alone suggest.

Now the humbling part: most people don't read every word. Nielsen Norman Group's analysis of a browser study found that on an average web page, people have time to read at most 28 per cent of the words, and 20 per cent is more likely. Its eye-tracking research describes scanning as the default, with thorough reading reserved for highly motivated readers. A document sent to someone personally may earn more of that motivated reading, but a time below the baseline usually means a skim, and that is normal, not an insult.

Reading a single visit

Individual visits are where the stories are, as long as you treat them as hypotheses rather than verdicts. These are common patterns, what they might mean and a sensible response:

Two of these deserve a closer look.

Long time on one page

A long dwell is the most tempting signal to over-read. Twelve minutes on the pricing page could mean "this is exactly what we need" or "this is twice our budget and I'm working out how to say so". It might also mean the reader stepped away, within the five-minute window. Use it to choose the topic of your follow-up, not to predict the answer.

Forwarded links

A forwarded link is often good news: your contact is showing the document to colleagues. With one named link per recipient (Breathier allows up to 50 links per document), a second device or an unexpected country on "Acme Ventures" is a fair sign the document has travelled. If the link asks for an email, remember that the address is typed, not verified, so a forwarded reader may simply type something else.

Reading across visits: the average chart

The average time on each page across visits shows the shape of attention to the whole document. It is good at answering questions such as: which pages does everyone skip? Where does interest drop off? Which page holds people longest?

Averages have a weakness, though. The NIST engineering statistics handbook notes that extreme values in the tails distort the mean, and that the median is the better guide to the typical value when such outliers exist. In document terms, one reader who left a tab open on page 6 can make page 6 look like the star of the show. With a handful of visits, open the individual ones before drawing conclusions from the chart.

Some practical rules:

• Compare like with like. A 500-word page will usually beat a 40-word slide. Judge each page against its own reading baseline, not against its neighbours.
• Filter by link. Investors and advisers, or this month's prospects and last month's, read differently. Filtering keeps each audience's pattern from blurring into the others.
• Wait for enough visits. With three visits, one person's habits dominate. Patterns worth acting on usually need more than a handful.

The caveats, in one place

Every figure in document analytics is a proxy for attention. These are the ways the proxy can drift from reality:

1. Idle tabs and still readers. Up to five minutes after the last movement can be counted, and a motionless reader can be undercounted.
2. Skimming. Time below the reading baseline usually means scanning, which is how most people read most things.
3. Forwarding. One link can have many readers; one visit is not one person.
4. Shared screens. A document projected in a meeting is one visit with many readers.
5. Downloads and printing. Once the file is downloaded, reading moves offline and out of view. Blocking downloads keeps reading in the browser, though anyone can still take a screenshot.
6. Unverified emails. The address on a visit is what the reader typed.
7. Automated checks. Business email systems check links; Microsoft's Safe Links scans URLs before delivery and again at the click. A visit with almost no reading time moments after you sent the email may be a scanner rather than a person.
8. Attention is not agreement. The numbers say where people looked, never what they thought.

Better than email open tracking, not magic

Email "read receipts" built on invisible images have become unreliable. Apple's Mail Privacy Protection downloads remote content in the background when a message is received, rather than when it is viewed, so an "open" may mean nothing more than delivery to someone's Apple Mail. Document analytics measure activity on the document itself, and requiring a visible tab and recent movement filters out much of the noise, though they remain a proxy. For more on what can and cannot be tracked, see our guide on how to see who opened your PDF and which pages they read.

Turning numbers into action

A simple routine keeps analytics useful rather than a compulsive refresh habit:

1. Make one link per recipient, named after them, so every visit has context.
2. Decide before you send what success looks like. For a proposal, perhaps the pricing and timeline pages read properly; for a pitch deck, the team and traction slides.
3. Wait a sensible time. A day or two for most business documents; people read on their schedule, not yours.
4. Read the visits, then the chart. Look for the patterns in the table above, against each page's reading baseline.
5. Follow up on the topic, not the timing. "Happy to go deeper on the rollout plan if useful" lands well. "I see you spent four minutes and twelve seconds on page nine" does not.
6. Improve the document. Pages that everyone skips are candidates for cutting; pages where readers stall may need to be clearer.
7. Close the link when you are done. Switching a link off closes it at once.

If you send pitch decks, our guide to sharing a pitch deck with investors applies this to fundraising, and why to send a link rather than an attachment covers the sending side.

Be fair to your readers

Readers of a Breathier link are told, on the link's page, that the sender can see that they opened it and how long they spend on each page, and a privacy page explains what is recorded. That openness matters. If you handle personal data under the GDPR, the European Commission's guidance says people must be told how their data is used in a concise, transparent way and in clear, plain language.

Beyond the law, there is courtesy: use what you learn to be more helpful, not to make people feel watched. If you would be uncomfortable explaining how you knew something, don't mention it. You'll find more guides like this in our resources hub.

FAQ

What is a good time on page for a document?

It depends on how much is on the page. Adults reading English non-fiction average about 238 words per minute, so a 300-word page takes a careful reader roughly a minute and a quarter, and a 40-word slide about 10 seconds. Compare each page's time with that baseline rather than with an arbitrary target.

Does a long time on a page mean the reader liked it?

Not necessarily. A long dwell shows attention, which can come from interest, doubt, confusion or simply an interruption. Use it to choose what to talk about in your follow-up, and let the reader tell you what they think.

Why does a visit show almost no reading time?

The reader may have opened the link and closed it straight away, glanced at it on a phone, or been interrupted. Reading time also stops when the tab is hidden or nobody has moved for five minutes. Automated link checks by email security systems are another possible explanation for a visit seconds after sending.

Can I tell if my document was forwarded?

Not with certainty, but there are strong hints. If you give each recipient their own named link, visits from new devices or countries on that link suggest it has been passed on. Email addresses typed at the gate are not verified, so they are not proof of identity.

What happens to the statistics if the reader downloads the file?

The visit records that the file was downloaded, but anything the reader does with the downloaded copy happens offline and cannot be measured. If page-by-page reading matters to you, turn downloading off for that link. That keeps reading in the browser, though it cannot prevent screenshots.

Is it fair to track how long people spend on each page?

It is fairest when readers know. Breathier tells readers on the link's page that the sender can see when they open it and the time spent on each page. Use the information to be more helpful, and avoid quoting exact timings back at people.

Sources

• Marc Brysbaert, "How many words do we read per minute? A review and meta-analysis of reading rate" (author's manuscript; published in the Journal of Memory and Language, 2019): 238 words per minute for silent reading of English non-fiction, a typical range of 175 to 300, and slower rates for second-language readers and older adults.
• Nielsen Norman Group: How Little Do Users Read?: people read at most 28 per cent of the words on an average web page, 20 per cent being more likely.
• Nielsen Norman Group: F-Shaped Pattern of Reading on the Web: scanning as the default and thorough reading when readers are highly motivated.
• MDN Web Docs: Page Visibility API: when browsers treat a page as hidden (background tab, minimised window, screen off).
• Google Analytics Help: Dimensions and metrics: user engagement defined as time a web page is in focus or an app is in the foreground.
• NIST/SEMATECH e-Handbook of Statistical Methods: Measures of location: extreme values distort the mean; the median is the better estimate when outliers exist.
• Apple: Mail Privacy Protection and Privacy: remote content downloaded in the background by default, limiting what senders learn about opens.
• Apple Mail User Guide: Protect Mail Activity: remote content downloaded privately when a message is received, instead of when it is viewed.
• Microsoft Learn: Safe Links overview: links in email scanned before delivery and checked again at the time of click.
• European Commission: What information must be given to individuals whose data is collected?: information must be concise, transparent and in clear, plain language.