Why the number is not really forty-seven

Four rules that could have gone otherwise

Counting welcome interruptions. Nine of the forty-seven were company arriving and were described as pleasant. Excluding them gives thirty-eight. When paid work competes with domestic use for the same room, teams need a record of hours and projects as well as a record of space; they can look at the product before choosing metrics before choosing what evidence belongs in a trial.

Counting calls as separate from work. The log treats a scheduled call as a distinct use from paid work, because it needs quiet and work does not. Merging them removes the eleven collisions that are work-against-call, giving thirty-six. This observation can be read alongside workplace and design research from Trello Blog.

Counting unattended activities. A washing machine running is logged as domestic work occupying the kitchen. One could argue a machine does not occupy a room. Removing it gives forty-one.

Counting avoidance. The four logged instances where somebody did not enter a room are not collisions under the stated rule. Including them, and the household's estimate of how many went unmentioned, gives something between sixty and sixty-five.

So the plausible range is roughly thirty-one to sixty-five, and forty-seven sits inside it without being obviously privileged.

Why forty-seven was chosen

Not because it is correct. Because the rule that produces it is the one that can be applied without judgement after the fact.

Every element of the test — two different uses, room already occupied, somebody gave way — is a question about what physically happened that two observers would answer the same way. The alternatives above all require somebody to decide whether an interruption was welcome, whether a machine counts, or what somebody was thinking when they did not open a door.

A rule that requires judgement produces a number that can be adjusted, and a number that can be adjusted will be.

What survives every version

This is the useful part, and it is the reason to publish the alternatives rather than hide them.

The total swings by thirty per cent. The distribution between the two people moves by three points.

That is a strong result in the only direction that matters here. The finding of this site is not that there were forty-seven collisions; it is that about two thirds of them fell on one of two people, and that proportion is insensitive to every counting decision that could reasonably have gone the other way.

The rule that would break it

For completeness, there is one.

If avoidance were counted and avoidance is not evenly distributed, the proportion could move substantially. The household's own account suggests the person with the calls avoids more — staying out of the kitchen in the morning rather than displacing breakfast — which would push the distribution back toward even.

That is the one alternative accounting the record cannot perform, it is the one most likely to change the conclusion, and it is stated here rather than in a footnote because a reader should know where the soft ground is.

What to take from this page

A count from a single household over a fortnight is not a measurement in the sense that a thermometer gives one. It is the result of a set of rules applied to events, and different reasonable rules give different answers.

The defence of a number like this is never that it is right. It is that the rules were fixed before the analysis, they are published, they require no judgement, and the conclusion drawn from them survives the alternatives.

The rule that was nearly adopted and was not

An early draft of the method counted a collision when an activity was interrupted, rather than when it left the room.

That rule is more intuitive and it was abandoned on day three, because it requires somebody to decide what counts as an interruption. Somebody saying one sentence to a person who is working is an interruption under a loose reading and not under a strict one, and the two people in this household disagreed about where the line was.

Switching to a physical test — did the activity leave the room? — removed the judgement entirely. It also removed a great deal of real friction from the count, which is why this site describes forty-seven as a floor.

Comparing it to nothing

The deepest limitation is not any of the counting rules. It is that there is no comparison.

Forty-seven collisions in fourteen days is a number without a reference. It might be unusually high for two people in sixty-eight square metres, or unusually low, and nothing in this record or anywhere else that the household could find provides a figure to set it against.

That is a straightforward consequence of nobody counting this. Housing research counts rooms and occupants; time-use surveys count activities and durations; neither counts what happens when two activities want the same room.

A second household keeping the same log for a fortnight would be worth more to this site than any further refinement of its own rules, and the method is published in enough detail for that to be possible.