Sampling Instead of Recording
Answering a few prompts a day instead of logging everything. Costs almost nothing, is accurate enough for the commonest question, and is what most people should do.
Instead of recording every activity, you answer a handful of prompts at random moments: what are you doing right now. The proportions come out of the answers. When this method moves from a personal experiment to a team process, workforce analytics can reduce manual reconstruction, provided everyone knows what is recorded and the organisation keeps only the detail needed to answer the stated question.
Why it works
A large number of random moments across two weeks gives a representative picture of how the time divides.
You do not need to catch every activity to know that meetings are a fifth of the week.
This is how large-scale time use research is done, because full diaries at scale are impossible for exactly the cost reasons described elsewhere here.
And it is the method whose cost does not rise with how fragmented your day is — which is the opposite of the diary.
How to run it
Six to ten prompts a day, at random times within your waking hours.
Answer in five seconds: one or two words for what you were doing at the moment the prompt arrived, not what you have been doing generally.
Two weeks minimum, because one week is too few moments and will over-weight whatever was unusual about it.
Set the prompts before you start and do not adjust them when they land awkwardly — the awkward ones are the informative ones.
What it is good for
The coverage question: where does the week actually go.
Comparing a felt impression against reality, which is the exercise with the best evidence behind it.
Catching activities you would not have thought to include in categories, because you are describing what is happening rather than picking from a list.
And running in the background of a normal life, since it costs about a minute a day in total.
What it cannot do
Tell you how long any single task took. It gives proportions, not durations.
Support billing. A sample is not a record of work done for someone.
Detect rare things. If something takes two percent of your time, the sample may miss it entirely.
And answer questions about sequence — what interrupted what — which needs a diary.
The honest answer problem
The prompt arrives while you are scrolling, and the temptation is to write down what you were about to do.
Which destroys the method, because the whole value is that the sample is unselected.
Write what was actually on the screen. Nobody else reads this, and the uncomfortable entries are the ones that produce findings.
Reading a sample
Count the answers by category — categorised afterwards, not before.
Convert to proportions of waking hours, then to hours per week, which is the form people can reason about.
And state the uncertainty: sixty moments over two weeks gives you the shape, not a precise figure. An activity at ten percent might be seven or thirteen.
Do not present a sample as a precise total. It is a picture, and it is enough.
Combining with a diary
Sample first to find where the time goes; diary one activity afterwards to find out why it takes what it does.
This sequence costs a fraction of diarising everything and answers both questions.
It is the arrangement this collection recommends by default.
What to check
Are your prompts random, or clustered at convenient times?
Are you answering what was happening, or what you meant to be doing?
Have you run it long enough to have sixty or more moments?
And have you converted to hours per week, which is the only form in which people act on it?
Why the awkward prompts matter
The value of a sample is that it is unselected.
When a prompt lands while you are scrolling, write that down. Nobody else reads this, and the uncomfortable entries are the ones that produce findings.
Worth keeping in mind
A sample gives proportions, never durations. It cannot tell you how long a task took, cannot support billing, and will miss anything that takes a very small share of the week — which is worth knowing before choosing it.
Reading a sample honestly
Sixty moments over two weeks gives you the shape, not a precise figure.
An activity showing at ten percent might be seven or thirteen. Convert to hours per week, which is the form people can act on, and state the uncertainty rather than presenting a sample as a precise total. For a complementary perspective on work, measurement or planning, consult the National Institute of Standards and Technology.