Workflow mapping from day-in-the-life and over-the-shoulder videos

The short answer: log four events per source — tool switches, manual data transfers, waits, and re-checks — then rank the steps by elapsed time rather than by how much the person complained about them. Eight to fifteen recordings across different seniorities and company sizes produce a usable map, because this genre is far denser than talking-head content.

The hardest thing to get out of a customer interview is a description of the boring part of their job. Not because they are hiding it, but because it has become invisible to them: the export, the reformat, the paste into the other system, the check that it landed. A recording of someone doing the work captures all of that whether they mention it or not.

What counts as this genre

The useful material goes by several names — day in the life, over the shoulder, screen share, live build, workflow walkthrough — and the common feature is that the camera or capture runs while real work happens rather than while work is described. Polished tutorials do not qualify, because the retakes remove exactly the evidence you want.

The tell is friction on screen. If nothing ever goes slightly wrong, you are watching a rehearsed demonstration. If a file opens in the wrong application, a tab is hunted for, or a value is retyped because copy did not carry the formatting, you are watching the real thing.

Log four events, ignore the narration

Resist summarising what the person says. The narration is the part they curated; the behaviour is the part they did not. Four event types carry almost all of the product-relevant information.

Re-checks are the most under-read of the four. When someone verifies the same value three times across a workflow, the cost is not the seconds — it is that they cannot trust the previous step. A product that removes the reason to re-check is frequently more valuable than one that makes the step faster.

Complaints point at the wrong step

Annoyance and cost are only loosely correlated. The step people complain about tends to be the one that is fiddly, unpredictable or embarrassing. The step that eats the day tends to be smooth, repetitive and unremarked, which is precisely why it never comes up in an interview.

Both are worth knowing, for different purposes: the complaint tells you what the marketing copy should name, because it is what the buyer recognises; the time sink tells you what the product should actually do. Building for the complaint alone produces a product people like and do not renew, which shows up later as the pattern in churn signals you can read before you have customers.

Sample across seniority and company size

The same nominal job has different bottlenecks at different scales. A solo operator’s workflow is dominated by context switching; a person on a five-person team is dominated by handoffs and waiting for someone else; someone in a large organisation is dominated by approvals and systems they cannot change.

A map built entirely from one of those three describes a segment, not a market, and it will look wrong to everyone outside it. Eight to fifteen recordings spread across the three is far more useful than thirty from one, and this is the same segmentation instinct as in defining an ICP from video research.

Tool switches are where products fit

Almost every workflow product lives at a seam: the point where data leaves one system and enters another by human hand. Mapping the switches therefore produces a list of candidate product shapes directly, ordered by how often each seam is crossed and how much reformatting it requires.

Not every seam is worth attacking — some exist because the two systems genuinely should not be coupled, and some are crossed twice a month. What separates the good candidates is frequency multiplied by transfer cost, and the resulting shortlist feeds directly into picking your first integrations from research.

Where people stall is where onboarding will stall

A recording shows exactly where a competent person hesitates: which setting they hunt for, which concept they explain wrong, which screen they back out of. Those hesitations transfer almost perfectly into your own product, because the confusion is about the domain rather than the interface.

Collecting them while mapping costs nothing extra and produces the skeleton of a first-run experience, which is the argument made at length in building an onboarding flow from where people stall. The questions people ask mid-recording are the same list that should seed your first help pages, per writing product docs from the questions people actually ask.

Turning the map into a scoped first version

A workflow map makes scoping arguments concrete. Instead of debating whether a feature is essential, you can point at the map and say this step consumes eleven minutes and is performed four times a day, while that one takes twenty seconds once a week. The first version should cover the former and ignore the latter, however loudly the latter is complained about.

The discipline of cutting to the few steps that dominate the map — and explicitly writing down what you are choosing not to touch — is the same one applied to build-along content in scoping a first version from build-along videos.

Treat every workaround as a specification

The most direct product signal in this genre is a homemade workaround: a spreadsheet with three tabs that exists only to reformat an export, a naming convention held together by discipline, a saved search someone rebuilds every Monday. Each one is a feature specification that a user already wrote, tested against reality, and decided was worth their own time to maintain.

Workarounds are better evidence than feature requests because they cost their author something. Nobody maintains a three-tab spreadsheet for a problem they merely find annoying, so the existence of the artefact proves both the need and a minimum willingness to pay measured in hours. Capturing the structure of the workaround — what it takes in, what it emits, what it protects against — gets you most of a product requirement for free.

The useful follow-up question is why the workaround has not been replaced by an existing tool. Sometimes no tool covers it; sometimes one does and the person does not know; sometimes the tool exists and is rejected for a specific reason. Only the first is an opening, and separating the three is the difference between a product and a marketing problem.

What a mapping pass costs

This genre is dense, so the corpus is small: eight to fifteen recordings is a full map. As of September 2026 that sits inside the Hobby plan at $19 a month with 25 videos and 2 projects; Pro at $59 covers 80 videos and 8 projects if you are mapping several roles, and Studio at $199 covers 250 videos, 20 projects and 3 seats. Every plan starts with a 7-day free trial — see the pricing page.

Closing thought

The most valuable minute in a workflow recording is usually one where nothing is said. Someone is exporting a file, renaming it, and pasting it somewhere else, and it has never once occurred to them to mention it.