Opened Versus Read: Which One Does Your Document Tracking Actually Know?

Everything in this field is sold on the word “read”. Know when your proposal is read. See who read it. The word is doing an enormous amount of work, and it is worth pulling apart, because the difference between what is measured and what is inferred is the difference between a tool you can rely on and one that will eventually embarrass you in front of a client.

There are three levels here, and only the first is a fact.

Level one: opened. This is measured.

A link was requested, a document was served, and it happened at a particular moment. This is as close to certain as anything in analytics gets — it is an event that either happened or did not.

The only qualifications on it are about who or what did the opening: automated traffic is filtered out, and reads from inside your own workspace are marked and kept separate from your client numbers. Past those two, “opened” is a fact you can build on.

Level two: read. This is inferred from time on screen.

Nothing observes reading. What is observed is that the document was the thing in front of the reader for a certain number of seconds — the count runs while the document is on screen and pauses when the tab is hidden, so a document left open in a background tab over lunch does not come back as a forty-seven-minute read.

That construction rules out the biggest lie in this category. It does not, and cannot, rule out the smaller one: there is no idle detection. A document visible on a second monitor while its reader takes a call accumulates time. So a duration is an upper bound on attention — at most this much, possibly much less.

Which is why the honest sentence about a nine-minute session is “this document was on screen for nine minutes”, not “they read it for nine minutes”. The gap between those two matters most when you are about to act on it. What a very short read actually means works the same distinction from the other end.

Level three: understood, persuaded, convinced. This is not knowable.

No behavioural signal reaches this level and none ever will. A person can read every word of a proposal and take nothing from it; another can skim it and grasp it completely because they have seen forty like it.

The reason to state something so obvious is that level three is exactly where people’s conclusions land. “They read it for nine minutes, so they must be seriously considering it” quietly travels from a measured fact through an inference to a claim about somebody’s state of mind, and nothing in the journey announces itself.

The specific things that are not recorded, and why it matters

It is easier to trust a measurement when you know what sits next to it and is deliberately absent:

• Where in the document they got to. There is no scroll depth and no section-by-section event. “They spent four minutes on the pricing section” is not a sentence this data can produce, and any tool that produces it for you is inferring from a scroll position — which measures where a window sat, not where eyes went.

• Where they came from. No referrer is recorded. You cannot tell whether a link was clicked in an email, a chat message or a calendar invitation.

• Who they are, unless they told you. Identity comes from being signed in, or from an address typed into an email gate — and the second is labelled self-reported everywhere it appears, because nobody verified it. The address you mailed the link to is not treated as an identity at all, because links get forwarded.

• Whether they printed, saved or forwarded it. A forward is inferred from a new-looking reader; nothing labels one.

How to phrase what you know

A practical habit that costs nothing: write the sentence out before you act on it, at the level of certainty the data supports. The three levels have three different verbs.

• Measured: “it was opened on Tuesday at 14:20, and again on Thursday.”

• Inferred: “it was on screen for about nine minutes across two visits, which for us is a long one.”

• Not knowable: “they liked it.” — delete.

If your plan only makes sense when you allow yourself the third sentence, the plan is running on hope rather than on evidence, and you should know that before you send the email.

Why the vocabulary is worth being fussy about

This is not pedantry for its own sake. A number that overstates itself is worse than no number, because it displaces the judgement you would otherwise have used.

It shows up in the interface too: a unique-reader count is printed as “≥ 12” rather than “12”, because a reader who blocks cookies looks new on every visit and the figure can only be a floor. A city is prefixed with “≈” because roughly one in four is wrong. A typed-in address carries a badge saying nobody checked it. Those marks exist so that a screenshot pasted into a client thread cannot say more than the data does.

The complete inventory — what is recorded, what is filtered and what deliberately does not exist — is on the features page. For the conclusions this all rules out, what document analytics can never tell you is the list.