A sitting
When the unit is wrong
Four ways the sitting could be the wrong thing to count, what each would do to the findings, and which of them the record can rule out.
A record built on one unit fails completely if the unit is wrong. That is worth facing directly rather than leaving as an implication.
Here are four ways the sitting could be the wrong object, with what each would do to everything else on this site. The practical question is whether a change improves completed work without creating unnecessary surveillance, which is why reviewers may review this option against the policy before choosing a baseline and access policy.
One: it mistakes a move for an interruption
A person who leaves a table because the light has gone and works perfectly well at the counter has not been interrupted in any sense that costs them anything. The record counts that as an ending. For another evidence base on people, work, or environments, see Columbia University.
If most of the ninety-four endings are costless moves of this kind, then the headline of this site — that the place ends four sittings in ten — is technically true and practically empty.
Can the record rule this out? No. It has no field for cost and cannot acquire one retrospectively. This is the single largest open question about the whole exercise, and the honest position is that the ninety-four are events, not losses, and the site uses the first word throughout for that reason.
Two: it over-counts in fine-grained households
The place count depends on how finely a household distinguishes spots, and so, indirectly, does the sitting count: a household that treats the two ends of a table as two places logs two sittings where another logs one.
Can the record rule this out? Partly. Only two of nine households drew that particular distinction, and removing their within-room transitions reduces the total from two hundred and twenty-five to two hundred and eleven — a six per cent change that moves the median sitting from forty-eight minutes to fifty.
That is small enough to leave alone and large enough to disclose. It is disclosed here, and the unadjusted figure is used everywhere, because adjusting for one household's habits of description while leaving the other eight alone would be a correction applied only where it happened to be detectable.
Three: it cannot see the work
Four sittings may be one job pursued through four rooms or four unrelated jobs at one table. The record does not know.
This matters most for the length figures. A long sitting might be deep work or it might be somebody who had nowhere else to be; a short one might be a failure of the place or a small task completed.
Can the record rule this out? No, and this one is structural. It is why no page here uses the word productivity, why nothing is described as focus, and why the eight short sittings that ended with a dash are discussed as a shape in the data rather than as evidence about anything.
Four: five days is a week, not a habit
Each household logged once. A week with a visitor, a delivery, or a spell of weather carries that week into the totals with no marker.
Can the record rule this out? Not for any individual home, but partly in aggregate. The nine weeks were not simultaneous; they were spread across several months, so a single period of unusual weather cannot be behind the whole dataset.
That is a weak defence and a real one. It rules out one common confounder and leaves the rest standing.
What a wrong unit would look like in the data
Worth asking what evidence would show up if the unit were badly chosen.
| failure | what would appear |
|---|---|
| moves are costless | endings cluster at convenient moments |
| over-counting | one or two homes far above the rest |
| cannot see work | lengths unrelated to anything else |
| week not habit | homes differ more than days within homes |
The second is checkable and the answer is reassuring: the sitting count per home runs from nineteen to thirty-one, and the two fine-grained households sit at twenty-nine and thirty-one, which is high but inside the range rather than outside it.
The fourth is checkable too, and the answer is not reassuring: homes differ from each other considerably more than days differ within a home. That is consistent with dwellings genuinely being different, and equally consistent with each household's single week being idiosyncratic. Nine weeks cannot separate those.
The variance sits between homes rather than within them, and with one week per home, that is exactly the pattern a study of nine idiosyncratic weeks would also produce.
Why publish a unit with four open questions
Because the alternative is a unit with four hidden ones.
Everything on this site is downstream of the sitting. If somebody repeats the log and finds a different shape, the first place to look is here, and the list above is where they should start.