Counting
Counting the Readings Carefully
Open counts are the only available evidence that a page was used, and almost every way of reading them is wrong in a predictable direction.
What nobody wrote down here, and what it cost
That the runbook saved somebody a morning
Nobody told the author
Six occasions, none reported back
That a note ended an argument
The author never heard
Both times from people who joined later
two things nobody wrote down · 8 occasions it cost somebody something
Two things, eight occasions, and the author heard about none of them.
Before adding another required page, managers can work-from-home monitoring solution and check whether the underlying bottleneck is missing knowledge, overloaded authors, or unclear ownership.
For a separate specialist reference on publishing and digital practice, consult W3C.
Four pages and the weeks each was opened, with what the raw count would have claimed
A page opened by nine different people
Eight weeks; the raw tool count was 46, inflated by one person's repeated visits
A page opened once a week by its own author
Eight weeks, and seven of them are the author checking something
A page linked from a dashboard everybody loads
Eight consecutive weeks, mostly incidental loads rather than readings
A page nobody opened
Zero, and this is the only unambiguous figure of the four
four documents · 24 weeks in which somebody opened one of them · the longest wait between writing and a reading was 32 weeks · one never read at all
Four pages, twenty-four weeks in which somebody opened one of them, and one never read at all. Three of the four counts mean something other than what they appear to mean. The fourth is the only one that can be read straight, which is a reasonable summary of what these numbers are worth. One team's own three quarters.
Tools generally record that a page was opened. That number is the only quantitative evidence available that any of this writing matters.
It is also wrong in several specific ways, and a team that starts reporting it without understanding them will draw confident conclusions that are false.
The four ways the number lies
Self-visits. Authors open their own pages constantly, to check something or to link to them. On a lightly-used page this can be most of the count.
Repeat visitors. One person opening something fifteen times while working through a procedure is one use, not fifteen.
Incidental loads. A page linked from somewhere everybody passes collects opens from people who did not read it.
Zero as ambiguous. A page with no opens may be useless or may be unfindable, and the number cannot distinguish them.
The reading that survives
Distinct people, over a period, excluding the author.
That is a much smaller number and it means something: this many different people needed this. Nine people in a quarter is a page earning its keep. Forty-six opens by four people is a page somebody is working through repeatedly, which is interesting for a different reason.
Most tools can produce distinct visitors. Almost nobody asks for it because the larger number is more flattering.
What the number is for
Two legitimate uses and one illegitimate one.
Telling an author that their work was used — the subject of the whole of this section, and the only feedback channel that scales.
Deciding what to write next, by noticing which of the existing pages people actually come to.
Not: comparing people. Which is covered below and is the thing that ruins it.
Why zero is the useful number
The fourth row is the only unambiguous figure in the panel and it is also the most actionable.
A page with no distinct readers after a quarter is telling you something. Either nobody can find it, in which case the title, the location or the links are wrong; or it answers a question nobody has, in which case writing more like it is a mistake.
Both of those are worth knowing and neither is visible without the count.
What goes wrong when it becomes a metric
Predictably: people optimise it.
Pages get linked from high-traffic places to collect opens. Titles become broader to catch more searches. Long pages get split into many short ones because each one then counts separately.
None of that improves anything and all of it is a rational response to being measured on the number. This is the subject of a later page in this section and it is worth avoiding rather than managing.
How to use it without that happening
Keep it private to the author, as feedback rather than as reporting.
A quarterly message saying "these pages of yours were opened by this many different people" is useful, carries no comparison, and cannot be gamed because nothing hangs on it.
The moment the same numbers appear in a shared dashboard with names beside them, the behaviour starts changing and the numbers stop meaning anything.
Reading the number correctly
The raw count is wrong in four specific directions, and knowing which is the difference between evidence and a confident mistake.
| What inflates it | What it conceals |
|---|---|
| The author's own visits | Whether distinct people came |
| One person opening it fifteen times | Whether anybody finished reading |
| Links from high-traffic pages | Whether the page was found or stumbled into |
| Nothing | Questions that stopped being asked at all |
Distinct people, over a period, excluding the author, is the figure that survives. It is much smaller than the headline number and it is the only one that means what it appears to mean.
The thing the count cannot see
What was prevented.
A page that stops a question being asked produces no reading, because the person who would have asked never needed to — they saw the title in a search and knew the answer existed, or a colleague linked it once and the matter was settled.
The tally of questions, kept before and after, catches some of this and open counts cannot. Where the two disagree, the tally is the better instrument, because it measures the thing anybody actually cared about.
The measurement the numbers cannot make
Open counts record arrivals. The thing most worth knowing is a departure: the question that stopped being asked.
A page that works does not necessarily get opened often. It gets mentioned once, linked twice, and then the question dies, because people learn the answer or learn that it exists. The count for such a page can be low while its effect is large.
The tally of questions, kept for a month before writing and a month after, catches exactly this and no instrumentation does. Where the two measurements disagree, the tally is the one to believe, because it tracks the thing anybody actually wanted to change.