The First Week of Data

The week is over and there is a record. What happens next determines whether the exercise was worth doing. When this method moves from a personal experiment to a team process, learn more about the workflow can reduce manual reconstruction, provided everyone knows what is recorded and the organisation keeps only the detail needed to answer the stated question.

Read before you count

Read the entries in order, as a narrative, before summing anything.

Most findings are in the sequence rather than the totals: what followed what, where the day fell apart, what was started three times.

Totals tell you proportions. Sequences tell you causes, and only one of those suggests a change.

Twenty minutes with the raw record, first.

The first questions to ask of it

Where did the day actually begin? Not when you sat down — when the first real work started.

What was the longest uninterrupted stretch, all week? For many people this number is shockingly small and is the most useful single figure in the record.

What happened immediately before each long stretch? There is usually a pattern, and it is repeatable.

What is in the record that you would not have predicted?

Discount the first days

The first two or three days are not representative: the observer effect is strongest at the start.

If you have only one week, note that the picture is tighter than reality.

This is the argument for two weeks rather than one, and it costs almost nothing extra with sampling.

Do not categorise yet

Categories applied at this stage are categories invented to fit what you already believe.

Read first, then decide what the useful divisions are, which will be different from the ones you would have chosen in advance.

The note on categories explains what goes wrong when this is done in the other order.

The comparison that matters

Write down, from memory, where you think the week went. Then compare it with the record.

Do this before reading the record in detail if you can — the estimate is contaminated once you have seen the data.

The gap is the finding. It is specific to you, it is usually large, and it is the one result in this subject with real evidence behind it.

What not to conclude

That a low figure for deep work means you are lazy. It means the week was structured a particular way, which is a different problem with different solutions.

That an unusual week is typical.

That a proportion is a problem. Twelve hours of meetings is only too much relative to something, and the something has to be stated.

And that you must act on everything. One change is the right number and has its own note.

Deciding whether to continue

If the question is answered, stop. That is success, not abandonment.

If it produced a new question, run a targeted investigation on that alone.

And if it produced nothing, say so plainly — that happens, and its own note covers what it usually means.

What to check

Did you read the record before summing it?

Do you know your longest uninterrupted stretch this week?

Did you write down your estimate before looking?

And have you decided whether this continues, or is it continuing by default?

The comparison to make first

Write your estimate of the week from memory before reading the record in detail.

Once you have seen the data the estimate is contaminated and the most informative comparison in the exercise is gone.

Worth keeping in mind

Discount the first two or three days. The observer effect is strongest at the start, so a single week gives a picture tighter than reality — which is the argument for two weeks rather than one.

In short

Read the record before summing it, and write your estimate before reading it.

Those two rules in that order extract most of what a first week of data has to offer.

The point

Read the entries in order as a narrative before summing anything.

Totals give proportions; sequences give causes, and only one of those suggests a change. For a complementary perspective on work, measurement or planning, consult Atlassian.