Interviewing as a Habit, Not a Project
One interview a week beats twenty interviews once a year, and it is easier to sustain.
Your company runs research properly. Twice a year, an agency does a study, delivers a deck of around sixty slides, and everyone agrees it was valuable. Six weeks later the deck is unopened and the team is making a decision about a checkout flow with no customer input at all, because the study did not cover checkout and the next one is in March.
Big-bang research fails the same way big-bang releases do. It is expensive, it takes a long time to produce, it arrives all at once, and it is stale before it is finished. Meanwhile the decisions that actually shape the product happen weekly, in small rooms, and they need small answers quickly.
At a minimum, weekly touchpoints with customers, by the team building the product, where they conduct small research activities in pursuit of a desired outcome.
Read that definition slowly, because each clause is doing work. Weekly: frequent enough that you always have something recent. By the team building the product: not a researcher who hands you a summary, because the value is largely in having heard it. Small activities: twenty five minutes, not a study. In pursuit of a desired outcome: focused, so the conversation goes somewhere rather than being a general chat.
Research as a project
- Runs twice a year, scoped in advance.
- Answers the questions you had six months ago.
- Delivered as a deck by someone outside the team.
- Recruitment takes weeks, so it only happens for big decisions.
- Engineers hear it second hand, if at all.
Discovery as a habit
- Runs every week, same slot.
- Answers the question you have this sprint.
- The people who will build it are in the room.
- Recruitment is automated, so the cost per conversation is near zero.
- A designer and an engineer attend and argue about what they heard.
The hardest part is not the interviewing. It is recruitment. Teams that try to schedule interviews on demand spend more effort chasing participants than talking to them, give up after three weeks, and conclude that continuous discovery: Talking to customers on a regular cadence rather than in occasional research bursts. does not work for their context. Solve recruitment once, structurally, and the habit sustains itself.
Four ways to make recruitment automatic
- Hook it to an event in the product
- After someone completes a first invoice, or cancels, or hits an error three times, show a small prompt offering a short call. You get relevant people at the moment the experience is fresh.
- Take a standing slice of support
- Ask support to forward one ticket a week that sounds interesting, and offer those people fifteen minutes. They already contacted you, so the yes rate is high.
- Ride an existing meeting
- Customer success runs check-ins constantly. Ask for the last ten minutes of one per week. Cheaper than any recruitment and the relationship already exists.
- Hold the slot regardless
- Put a recurring hour on the calendar for the whole trio. If nobody is booked, use it to watch session recordings or read tickets together. Protecting the time is what keeps the habit alive through bad weeks.
Worked example
Ledgerline again, six months in
The team put a one line prompt at the end of the monthly invoicing run: finished your run, got fifteen minutes to tell us how it went? Roughly one person a day accepted. The PM, a designer and one engineer took turns, one call each per week, twenty five minutes, no deck. Three months in, the engineer noticed something nobody had planned to look for: four different customers described checking the invoice list against their bank statement line by line before sending. That was not a feature request and it had never appeared in a ticket. It became the biggest thing the team worked on that year. It surfaced only because somebody who could build things was in the room every week with no agenda except listening.
A year of each approach
Keeping the habit from decaying
- Rotate who runs the call, including engineers. A team where only the PM interviews has one person's interpretation of every customer.
- Cap it at twenty five minutes. Long interviews get scheduled less often, and the marginal value after twenty minutes is small.
- Write the notes within the hour. Interviews you have not written up stop being evidence within a few days and become vibes.
- Bring one open question from the current sprint, and let the rest of the conversation wander. A completely scripted call finds only what you already suspected.
- Share a two line summary in the team channel every week. Visibility is what stops the slot from being the first thing sacrificed when a release is late.
You will hear the objection that one interview a week is too small a sample to decide anything. It is, and that is not what it is for. Weekly interviews are not a substitute for quantitative evidence. They are the thing that tells you which numbers to go and look at, and that stops you from being confidently wrong about how people actually work. The sample builds up over months, and a team that has done forty of these has a far better model of its customers than a team holding a sixty slide deck from March.
Start with four weeks. Not a programme, not a proposal to leadership. One hour a week for four weeks, with one engineer, and a two line summary each time. If after four weeks nobody has learned anything that changed a decision, stop and say so publicly. In practice that almost never happens.
Quick check
Which failure most often kills a team's attempt at weekly customer interviews?
The takeaway
Discovery works as a weekly habit run by the team that builds the product, and recruitment is the constraint to solve first.
Try this tomorrow
Book a recurring 45 minute slot this week with one engineer and one designer, and set up a single automatic source of participants (a post-task prompt, or one forwarded support ticket a week). Hold the slot even when nobody is booked.
Answer the check above, then bank the day.
Where this comes from
- Continuous Discovery Habits, Teresa Torres
- Inspired, second edition, Marty Cagan
- The Mom Test, Rob Fitzpatrick
Discovery is one of six tracks. These lessons summarise and build on the work above, they do not reproduce it. Buy the books, they are better.