Cheaper Alternatives to Desktop Capture

Before installing an agent, there are four methods that answer most of the same questions. Documenting why they were rejected is required in several jurisdictions anyway.

Interviews

Ask the people who do the task to describe it.

Fast, cheap, and it gives you the reasons that no capture provides.

Weak on durations, which people estimate badly, and on the steps they have stopped noticing.

Best for: understanding why, finding the workarounds, identifying what the systems do not record.

Do this first, always. It frequently answers the question and always improves whatever comes next.

Observation

Sit with someone and watch, taking notes.

Gives sequence and approximate duration, plus the conversation about why.

Has the same observation effect as capture and is more transparent about it, which people generally find easier.

Best for: characterising a task well enough to model it.

A day of observation with three people frequently produces the same model as a two-week capture study.

Self-recorded logs

Participants record their own steps, on a simple form or a tally.

Poor precision, good coverage of the invisible, since people record the paper note and the phone call.

Burdensome, and it degrades after a few days.

Best for: counting the frequency of a specific interruption or manual step.

Better system logging

The most underused option.

Adding start and end timestamps separates working time from waiting.

Adding lifecycle states reveals suspension.

Adding a reason code on rejection converts a mystery into a distribution.

Enabling logging that is off by default, which is a configuration change.

This closes a large part of the gap with no monitoring dimension at all, and it is almost always cheaper than a capture study.

Where capture is genuinely better

Click-level sequences for automation, where the specification must be exact.

Precise durations of short steps, where estimates are unusable.

Application switching frequency, which nobody can self-report accurately.

Copy and paste between systems, which is the clearest automation signal and is invisible everywhere else.

Four narrow cases. Outside them, one of the alternatives above is usually adequate.

Recording the decision

Which alternatives were considered.

Why each was insufficient, specifically.

What the additional precision is worth.

What safeguards were added to compensate for the intrusion.

Recorded at the time, because a proportionality assessment reconstructed two years later during a dispute is worth very little.

Recording why each was rejected

Required in several jurisdictions and useful as a scoping discipline.

Interviews: what they could not tell you, specifically.

Observation: why the precision was insufficient, with a number if possible.

Self-recorded logs: why the burden or the accuracy made them unusable.

Better logging: why it was unavailable or too slow, which is sometimes an honest answer and sometimes a preference for the faster route.

Written at the time, because a proportionality assessment reconstructed later during a dispute persuades nobody.

Doing interviews first regardless

Even where capture is justified, the interviews come first.

They tell you what to look for, which narrows the capture scope substantially.

They tell you what the systems do not record, which is the gap to target.

They surface the workarounds, which observed work conceals.

They give you the reasons, which no capture provides and which every finding needs.

A capture study designed after interviews is shorter, narrower and produces a better model than one designed from a process diagram.

What to try first

Four cheaper approaches that answer many of the same questions.

Extend the system logs. Frequently the missing events exist and are not extracted, or can be enabled by configuration.

Instrument the desktop tools themselves, where they have their own logging.

Observation. Sitting with three people for a morning answers most "what happens between events" questions, with consent and no capture.

Ask. A structured interview with the people doing the work produces the workaround list quickly and accurately.

Try these before proposing capture, because if they answer the question the proportionality case for capture disappears.