Video research vs customer interviews: what each one is for

The short answer: use video research to learn how the work is actually done and what people struggle with, then use a small number of interviews to learn what video cannot show — budget, authority, urgency and how a real person reacts to your idea. Video first, interviews second, and neither on its own.

Founders are told to talk to customers, and the advice is sound. The trouble is the first ten conversations. Before you know the vocabulary of a field, you ask naive questions, get polite answers, and spend most of each call being educated by someone who has better things to do. Much of what those early calls teach is available in public, already recorded, from people demonstrating their work on camera.

Observed behaviour and reported behaviour

The core difference is simple. An interview gives you reported behaviour: what someone says they do, filtered through memory and the wish to sound competent. A workflow video gives you observed behaviour: what someone did, in order, with the awkward parts left in.

Reported behaviour drops the steps people have stopped noticing. Ask an operations manager how they produce a weekly report and you will hear three steps. Watch one do it and you will count eleven, including two exports, a manual fix to a date format and a message to a colleague asking for a number. The eight missing steps are where products come from.

Observed behaviour, for its part, cannot be questioned. You cannot ask the creator why they did it that way, what they tried before, or whether they would pay to stop. The video is a one-way record.

Side by side

Each method has its own bias

Interview bias is well known. People are kind to the person in front of them, they predict their own future behaviour badly, and a hypothetical “would you use this?” draws a yes that costs nothing to give. Good interviewers work around this by asking about the past rather than the future.

Video bias is less discussed and just as real. People who make videos are not a random sample. They skew toward the skilled, the tool-curious and those with something to sell, and a meaningful share of content is sponsored. Popular videos over-represent problems that are interesting to watch being solved.

Sorting independent practitioners from people with a product to push is the first correction to apply to any video corpus; the method is in vendor content versus independent creators. The broader set of ways an AI reading of video can mislead is in failure modes of AI video research.

A sequence that uses both

The most efficient order is to let video do the broad, cheap learning and reserve conversations for the narrow, expensive questions.

Step one: build the map from video. Twenty to thirty videos from unrelated practitioners, synthesised into the workflow, the tools, the recurring complaints and the jobs people are trying to get done. At the end you should be able to describe a typical week in the role without notes. The output of this stage in job-statement form is described in writing jobs-to-be-done statements from video research.

Step two: list what the map cannot tell you. This is the step people skip. Write down the open questions explicitly: who pays, what it replaces in the budget, how urgent the problem is compared with the other ten on their list, what they tried and abandoned. That list is the interview guide.

Step three: interview against the list. Because you already know the workflow, you can skip the orientation and go directly to the gaps. Turning a synthesis into a set of questions is covered in building a customer interview script from video research.

Step four: reconcile. Where interviews contradict the video synthesis, the contradiction is the finding. Usually it means ordinary practitioners differ from the ones who publish, which tells you something specific about who your first customers are.

When interviews should come first

Video-first is the default, with clear exceptions. If the field is one where practitioners do not publish — regulated work, internal enterprise processes, anything covered by confidentiality — there may be little to watch, and conversations are the only route in. The options in that situation are set out in what to do when your niche has few videos.

Interviews also come first when you already have warm access. If five people in the target role will take your call this week, take the calls, and use video afterwards to check whether what they told you generalises beyond your own network.

And for the final question — will this person pay — neither research method is sufficient. The evidence that settles it is a commitment: a deposit, a signed pilot, a card on file. Research narrows the field to ideas worth asking that question about, which is the job of validating a SaaS idea without surveys.

What a combined finding looks like

Take a founder looking at scheduling for mobile dog groomers. The video pass shows a consistent picture: bookings arrive by text message, the route for the day is worked out by hand the night before, and cancellations are handled with a flurry of calls. Every creator in the corpus complains about the route planning.

On that evidence alone, route optimisation looks like the product. Five conversations then add what the footage could not. Three of the five groomers say the route is annoying but takes ten minutes. What costs them real money is the late cancellation, which leaves a paid slot empty. Two mention they already tried a general booking tool and dropped it because clients would not use the app.

Neither source was wrong. Video correctly showed the visible friction. Interviews showed which friction has a price attached and which obvious solution has already failed. The product that comes out of both — filling cancelled slots by text, with no client app — is one that neither method would have reached alone.

What each one costs

Interviews are cheap in money and expensive in calendar time. Finding people, scheduling, the call itself and writing it up can stretch ten conversations across a month. Video research inverts that: a small subscription and an afternoon. As of October 2026, Hobby is $19 a month with 25 videos and 2 projects, Pro is $59 with 80 videos and 8 projects, and Studio is $199 with 250 videos, 20 projects and 3 seats, each with a 7-day free trial — see the pricing page.

Closing thought

The debate about which method is better mostly comes from treating them as substitutes. They are consecutive. Watch first so that you know what to ask, then ask the things that cannot be watched.