Why we ask people instead of measuring them

There is an obvious objection to building a process map out of interviews. People are unreliable narrators of their own time. They round, they forget, and they describe the good version of a process rather than the usual one. Software that watches the screen does none of that. So why ask?

Because the two methods answer different questions, and only one of them is the question an automation project needs answered.

What tracking measures

Time-tracking tells you where hours went, with an accuracy no human estimate can match. Four hours in the CRM, ninety minutes in email, forty minutes in a spreadsheet. Every number is real.

What it cannot tell you is what any of it was for. Ninety minutes in email is not a process. It is a container that might hold six processes, most of them not worth automating. To get from the container to the process you have to ask somebody, which puts you back where you started, only later and after buying software.

Automation does not act on hours. It acts on a trigger, a sequence of steps, a set of tools and a handoff. Tracking sees none of those. It sees the shadow they cast on a clock.

What asking gets wrong, and by how much

The honest version is that estimates are soft. Someone who says a quote takes twenty minutes might mean fifteen or thirty, and the error does not cancel out neatly across a team.

Two things keep this survivable.

The first is that the interview asks about yesterday rather than about typical weeks. Recalling a specific day is a much easier memory task than averaging an abstraction, and it produces the interruptions and the false starts along with the work.

The second is that the arithmetic is deliberately blunt. Frequency times duration, priced at one blended rate, with self-reported ad-hoc time discounted by half on its own line. A soft input does not become firm by being multiplied through a more sophisticated model. It becomes a firm-looking output, which is worse, because nobody can see where to push back.

What only asking can get

Three things surface in an interview that no observation tool would produce.

The broken version. Every process has a version where a field is missing or the customer replies late, and it often costs more than the version that works. People describe it readily when asked what goes wrong. A screen recorder captures it as an ordinary afternoon.

The waiting. The gap between one person finishing and the next picking it up is invisible to per-person tracking, because during the gap nobody is doing anything. It is frequently the largest number in the map. That is the subject of [the cost of handoffs](/guides/the-cost-of-handoffs).

The reason. Why a step exists at all. A good share of the automation candidates in any company are steps that made sense in response to a problem that no longer happens. Nobody automates those once they hear the reason out loud. They delete them.

The part that makes the number smaller

Because the input is an estimate, the output is an estimate, and the report says so under the headline figure rather than in an appendix. A map built this way is good enough to choose what to automate first and to price the work. It is not good enough to be a billing record and should not be presented as one.

If you want a defensible measurement of hours, buy tracking. If you want to know which work to remove, ask the people doing it.