Industry
Part of What a customer interview can measure, and what it never will
Customer interviews framework: plan the round, not the call
A customer interviews framework for a whole round: the decision first, a sample frame, a stopping rule, batches and the four outcomes that close it.
A framework for interviewing usually means a running order for one conversation. That is the least valuable part. One conversation goes well or badly on its own, while a round of them either answers a question or does not, and the difference is decided before the first call.
A round has a question, a sample, a stopping point and a decision waiting on the other side. The framework below plans all four, in the order they have to be settled.
What to take away
- A round exists to answer one question and feed one decision. If you cannot name the decision, do not book the calls.
- Design the sample and the stopping rule before you speak to anyone. Both become unarguable once results start arriving.
- Run rounds in small batches with a write-up between, so what you learn can change who you call next.
- A round ends with a dated change or a dated confirmation, and anything else was encouragement.
Start with the decision, then the six fields
Name the decision before the first call, because it is the field that kills bad rounds. If both possible answers lead to the same next step, the round is entertainment, and it will still take three weeks.
Fill the six fields in before anything is scheduled. It takes twenty minutes and it is the whole framework.
Six fields before booking
- The questionone thing to establish
- The decisionwhat changes on the answer
- The sample framewho counts, where found
- The screenertwo or three factual questions
- The stopping rulewhat says finished
- The write-upwhere notes go, what shape
| Field | What goes in it | Failing version |
|---|---|---|
| The question | One thing this round is meant to establish | "Learn more about our users" |
| The decision | What you will do differently depending on the answer | Nothing, or something you would do anyway |
| The sample frame | Who counts, and where you will find them | "Anyone in the industry" |
| The screener | Two or three factual questions that admit or exclude | Asking whether they are interested |
| The stopping rule | What tells you the round is finished | A target number of calls |
| The write-up | Where notes go, and in what shape | Separate documents nobody rereads |
Build the sample frame and the screener
The sample frame says who counts and where you will find them. A screener then turns that frame into two or three factual questions that admit or exclude a person before you meet.
Screener questions: good vs bad
Good screening questions
- Checkable
- Which tool they use
- Frequency
- How often last month
- Role
- Whether they hold the role
- Buying
- Who they buy from
Bad screening questions
- Checkable
- Whether topic matters to them
- Frequency
- How interested they are
- Role
- Whether they would consider it
- Buying
- Whether they like the idea
Good screening questions have an answer you can check: which tool they use, how often something happened last month, whether they hold a particular role, who they buy from. Asking whether the topic matters to them is one of the mistakes before the call, because everyone answers yes when a stranger asks.
The wording deserves more care than a screener usually gets, because a screener is a short questionnaire. The Census Bureau's questionnaire standards say pretesting is critical to finding problems with question content and order effects, so read the questions aloud to someone outside the project first.
Keep the people you screened out. A list of those who looked right and turned out not to have the problem is a description of your boundary, and boundaries are hard to see any other way.
Set the stopping rule before the first call
The stopping rule is not a number of calls. It is about what the calls are producing, and it is written down before anyone is booked.
When to stop the round
Two or three calls in a row change nothing?
stop: saturation for this segment
keep going, write up, predict next
Stop when two or three conversations in a row change nothing in your written problem statement and you can predict what the next person will say about their current method.
That is saturation for this segment, and the wording matters: for this segment. A round of people you could reach easily saturates fast and tells you nothing about anyone else.
Two counts tell the two apart. Count how many of the round arrived by the same route, such as one community, one employer or one introduction chain. Then count how many share the same setup and the same constraints.
When nearly all of them do, you have saturated your address book, not a segment. The counts worth keeping are those two, tracked round by round.
Continue past that point only if the sample frame changes. A new segment is a new round, with its own question and its own screener.
Run the round in batches
Book three or four calls, then stop and write them up. Then decide who to call next.
Batches, not a schedule
- Book three or four conversations
- Stop and write them up
- Decide who to call next
- Askdid the statement change?
- Askwho are we missing?
Booking twelve in advance locks the sample to what you believed on the day you booked. Half the value of early conversations is that they tell you the wrong people are on your list, and a full calendar makes that discovery useless.
Between batches, ask two questions:
- Did anything change in the written problem statement?
- Is there now a kind of person we should be talking to and are not?
Those two are the reason to run in batches at all. If both answers are no for two batches running, the stopping rule is close.
Keep a ledger, one row per conversation
The write-up is a ledger with three columns, one row per conversation, never merged. Name the columns before the first call, because a grid designed after the fact records only what you already believe.
- What they did.The last time the problem occurred, in the person's own words, with your interpretation kept out.
- What it means for the question.The one thing the round was meant to establish, and whether this row supports it, contradicts it, or says nothing.
- What changes.The edit to the written problem statement or to the next round's screener, or the word "nothing".
The write-up shape
- Saidwhat the person reported, their words
- Observedwhat you saw happen
- Concludedyour reading, kept separate
- How you got to this person
- What surprised you
Add two fields beside them: how you got to this person, and what surprised you. The first tells you whether your evidence is about a market or about your address book. The second is the only field containing anything you could not have written in advance.
Read the what-they-did column before the what-it-means column. What someone actually said stays put, while your reading of it changes with every new batch.
What the round hands over
A round should end with a written change or a written confirmation, dated. Only four outcomes count, and each is worth writing down on the day it appears.
- The problem statement changes.The round showed the problem is different, narrower, or belongs to someone else.
- The sample changes.The round showed you are calling the wrong people, so the next round gets a new screener.
- The problem statement holds.The same account came back from people who reached you by different routes, and the decision moves forward.
- The question closes.The problem is real but nobody is acting on it, and you stop spending rounds on it.
What a round hands over
- Narrower problem statement, named segment
- List of people who looked right but are not
- Who authorizes a purchase, what precedent looks like
- The vocabulary customers actually use
A round that produced none of these produced encouragement. Run the next one with a different sample rather than more calls, since more of the same sample cannot fix a sampling problem.
Where the round stops
Interviews can establish what happens today, what it costs, and who decides. They cannot establish that anyone will pay or count anything, and no number of rounds changes that, because a description is not a transaction.
Counting is market research, which the SBA describes in its guidance on market research and competitive analysis as helping you find customers for your business. See market validation for the counting work interviews cannot do.
The next evidence comes from a specific offer at a specific price that somebody can refuse, which is a different apparatus with a different unit of evidence. Minimum viable offers covers keeping the first one small.
Common questions
How many rounds does a project need?
One per segment you are serious about, plus a repeat whenever the problem statement has changed enough that your screener would now admit different people. Usually two or three, not ten.
Can one round answer two questions?
Only if the same conversation produces evidence for both without the questions interfering. Usually the second changes how you ask the first, and you end up with two half-answers.
What if I cannot fill the sample frame?
That is a finding, and an important one. Being unable to reach twenty of the people you claim to serve is a distribution problem sitting in front of everything else, and it does not improve after you build.
Do I need a formal discussion guide?
A short one: three or four openings and a list of things you want to reach, not a script. What makes a conversation productive is following the friction wherever it goes, and the guide's job is to stop you forgetting the buyer question, not to fix the order. Customer interviews covers the moves inside the room.
