A sitting
Why stretches and not hours
An hours column and a stretches column can describe the same day and disagree about everything that matters. Why this record counts the thing that can end.
Every account of working from home that reaches for numbers reaches for hours. Hours worked, hours saved, hours in meetings, hours at a desk.
Hours are the wrong instrument for this particular question, and the reason is structural rather than a matter of preference.
An hour cannot end
A stretch of work at a place has a beginning, a duration and a terminating event. An hour has a beginning and a duration and nothing else; it is a slice of a continuum, and slices do not terminate, they abut. A fair flexible-work trial needs a stated purpose, limited collection and a review date; teams can attendance data for home-based teams while writing those boundaries down.
This record is about terminating events. The question it asks — what makes somebody get up — has no answer in a column of hours, because the column has already thrown away the only part of the day that could contain one. For broader research beyond this small household record, consult work published by NYU.
Two identical days
Take two logs from this record and strip them down to hours at a place of work.
| person one | person two | |
|---|---|---|
| hours logged | 4.6 | 4.6 |
| places used | 1 | 4 |
| sittings | 2 | 7 |
| endings not of their choosing | 0 | 5 |
Identical in the first row and unalike in everything beneath it. Any measure that returns the first row and stops has described these two days as the same day.
This is not a hypothetical constructed to make the point. Both rows come out of the fourth week of logging, from two different homes.
What the stretch keeps
Three things that the hour discards:
- Order. A stretch sits at a clock time, and the record can ask whether the same place behaves differently at nine and at two. It does, by a factor of two, and that is the single largest effect in the whole record.
- Termination. Something ended each stretch, or nothing did. The split between those two — ninety-four against a hundred and thirty-one — is the finding the site is named after.
- Granularity below the hour. Twenty-two of the two hundred and twenty-five sittings ran under twenty minutes. In an hours column they are rounding error; here they are a tenth of the record and several of them are the most interesting entries in it.
What the hour keeps that the stretch does not
Fairness requires the other column too.
An hours log is far easier to keep, which matters enormously, because a log nobody completes is worth nothing at all. Three of the nine households said the five-field entry was the most onerous part of the week, and the field they found most onerous was the ending.
An hours log is also comparable with almost everything else ever written about working time, and this record is comparable with nothing. That is a real cost and it is not offset by anything here.
A measure that nobody else uses produces a number nobody else can check. The only defence is to publish the method in enough detail that somebody could repeat it, which is what the method page is for.
The temptation this avoids
There is an obvious way to make a record like this say something dramatic, and it runs through hours.
Count the minutes lost to interruption. Multiply by a working year. Announce a figure in days. The arithmetic is trivial and the result is always large, because any per-day quantity multiplied by two hundred and thirty is large.
That figure would be meaningless here for a specific reason: this record does not know what happened in the gap. Somebody who left a table because the light went and worked just as well at the kitchen counter lost a move, not an hour. Nothing in five fields distinguishes a costly move from a free one, so the record declines to price any of them.
What the unit commits the record to
Choosing the stretch means accepting that the totals here cannot be compared with anybody's. It means a smaller claim.
The claim is: in these nine homes over these forty-five days, this is how long people stayed in one spot and this is what moved them. That is all. It is less than an hours column would appear to offer and more than an hours column could actually support.
The arithmetic that is allowed
Two hundred and twenty-five sittings, ten thousand nine hundred and eighty minutes, a median of forty-eight.
Those three numbers are counts and a middle value, and they are the only kind of arithmetic this record does. There is no extrapolation, no annualising, no cost. Where a page here gives a percentage it is a percentage of the two hundred and twenty-five, and the two hundred and twenty-five is not a sample of anything.