What PDF analytics actually tell you

Most document analytics stop at a number that tells you almost nothing: views.

A view means a browser loaded the first page. It does not distinguish the client who read all thirty-eight pages twice from the person who opened your proposal on a phone, saw the cover, and closed it. Both are one view. If you have ever sent a proposal and felt reassured that it was “opened three times”, you were reading a number that could not support the feeling.

Here is what is actually worth looking at, and what each thing can and cannot tell you.

Read depth is the one that matters

Read depth is how far through the document a reader got. Page 4 of 38, or page 38 of 38.

It is the most honest signal a document produces, because it is the hardest to fake by accident. Nobody reaches page 30 without meaning to. When you are sending the same proposal to five prospects, the one who reached the end is a different conversation to the one who stopped on the pricing page — and you now know which is which before you follow up.

It also changes what you write next. If everyone stops at the same page, that page is the problem. Not a hunch — a number with a page count attached.

Drop-off tells you where the document fails

Read depth across all your readers, laid out per page, is a drop-off curve. There is usually one cliff, and it is usually somewhere specific: the page where the density jumps, the page after the pricing, the page where you started explaining the process instead of the outcome.

This is the most actionable thing in the whole set, and almost nobody looks at it, because most tools do not measure per page.

Re-reads point at what mattered

The page people go back to is different from the page they spent longest on. Going back is a deliberate act. In a technical document it is usually the specification table; in a proposal it is usually the price.

That single fact — most re-read page — is often the fastest read on what a document is really being used for.

Time is useful for comparison, not for truth

Time on a page is measured while the page is on screen. That means it counts a tab left open over lunch.

Treat it as a comparison between pages in the same document — page 12 held people three times as long as page 11 — and it is genuinely useful. Treat it as “this client spent nine minutes considering our proposal” and you are building a story on a browser tab.

What analytics cannot tell you

This is the part that gets glossed over, so let us be blunt about it.

Who someone is. A link that anyone can open tells you a document was read, not who read it. If you need a name, you have to ask for one — an email gate before the document, or a paywall where checkout captures an address. Both are trade-offs: every gate costs you readers. A gate is a decision about which you want more, reach or names.

Whether they agreed with it. Reaching the last page is not consent. It is attention.

What happened outside the browser. Someone who downloads the PDF and reads it on a train is invisible from that moment on. That is one of the quieter arguments for sharing a document as a link rather than an attachment: an attachment stops reporting the instant it is delivered.

Anything from before you started. Analytics are not retroactive. There is no history for a document that has been emailed around for six months.

The retention limit, said plainly

On Living Page, reader events are kept for three months.

That is not an oversight, it is the trade that makes per-page analytics viable: the events go into an analytics engine designed for high-volume time-series data with a rolling window, which is what keeps per-page tracking fast and free of per-document cost. Storing every page-turn of every document forever in a normal database is how document analytics turn into a line item.

What it means practically: a quarterly review is comfortably inside the window, and a year-on-year comparison is not. If you need the numbers to outlive three months, capture them in a report while they are live.

What to actually do with this

If you are sending client work:

1. Look at read depth, not views. It is the only number that separates a reader from an opener.
2. Find the cliff in the drop-off. Fix that page before you write anything new.
3. Note the most re-read page. That is what your reader came for — lead with it next time.
4. Use time only to compare pages within one document.
5. Decide deliberately whether you want names. If you do, gate it and accept the drop in readers. If you do not, stop pretending the numbers will tell you anyway.

The uncomfortable summary is that a document you emailed gives you none of this, and a document you shared as a link gives you all of it — which is a stronger argument for changing how you send work than any feature list.