Customer discovery interviews, and how many are enough
B2B customer discovery interviews: the sample size the research converges on, why your segments change that number, and the questions that get honest answers.
Read
The percentage next to a deal stage is a default an administrator configured once, and the rep who owns the deal can usually edit it. Your review exists because that number cannot be trusted on its own.
Most pipeline review meetings are a reading exercise. Somebody shares the CRM, the team walks the list from largest to smallest, each rep narrates what they already wrote in the notes field, and an hour later nothing has been decided. The pipeline is the same pipeline. The forecast is the same forecast. Everyone agrees it was useful.
A review that earns its hour does one thing instead: it changes the state of deals. Some move forward with a named next step, some get pushed to a later month, and some get killed. If no deal changed category by the end, the meeting was a status report wearing a meeting's clothes.
Start here, because it is the most common source of self deception in a review. When a deal sits at a stage and the CRM shows a percentage next to it, that percentage is almost always a default that an administrator configured once, not a judgement about this deal.
Salesforce publishes its own default mapping, and it is worth seeing written out. In the documented default stage to forecast category mapping, Prospecting and Qualification are both 10 percent, Needs Analysis is 20 percent, Value Proposition is 50 percent, Id Decision Makers is 60 percent, Perception Analysis is 70 percent, Proposal/Price Quote is 75 percent, and Negotiation/Review is 90 percent. Closed Won is 100 percent and maps to Closed, while Closed Lost is 0 percent and maps to Omitted, which the documentation notes is "not included in forecasts".
Read that list again as a rep would experience it. Moving a deal from Needs Analysis to Value Proposition raises its apparent probability from 20 percent to 50 percent. Nothing happened in the account. A picklist changed. The same documentation also states that "Users can edit the Forecast Category field on opportunities that they own, overriding the default stage-to-category mapping", and that in Lightning Experience "all forecast categories are available for all stages". So the category is editable by the person with the most reason to be optimistic about it.
None of that is a criticism of the tool. The defaults have to be something, and a configurable mapping is the right design. The mistake is treating the output as a measurement. Your review exists precisely because that number cannot be trusted on its own.
Replace the narration with a single demand: what did the buyer do that you did not ask them to do. Buyer initiated actions are the cheapest honest signal available, because they cost the buyer something.
A deal with none of these is not a late stage deal, whatever the stage field says. It is an interested conversation, and interested conversations are where the forecast goes to die. This is the same failure the five places deals die covers from the other direction, and it is worth reading alongside: why B2B deals stall.
| What the rep says | What the stage implies | What to ask for in the room |
|---|---|---|
| "They are very interested" | Qualified, mid pipeline | Name the last thing they did unprompted |
| "Waiting on their legal" | Negotiation, near close | The name of the lawyer and the date it was sent |
| "Budget is approved" | High probability | Who approved it, and for which quarter |
| "Champion loves it" | Multi threaded | Who else has said anything, in writing |
| "Slipping to next month" | Still live, just late | What changed since the last date, and why this date is different |
The last row is the one that protects the forecast. A date that moves once is normal. A date that moves three times is a deal that is not real yet, and the honest action is to push it out of the committed number rather than to re-promise it. Slippage is measurable and it is the input most teams ignore, which is the argument in how to forecast B2B sales accurately.
These are two meetings and they fail when merged. The commit conversation is short, unemotional and about categories: what is in the number, what came out, what the total now is. The coaching conversation is about how to advance a specific deal, and it needs time, candour and usually fewer people in the room.
Run the commit pass on the whole pipeline first, fast, with no problem solving allowed. Then pick two or three deals for real inspection. Trying to do both at once is why reviews run long and why reps learn to perform rather than disclose.
Salesforce's documentation on rollup methods is a useful mental model for the first pass: with cumulative columns, the Best Case Forecast column combines Best Case, Commit and Closed, while single category rollups show each category on its own. Knowing which view you are looking at matters, because a cumulative number and a single category number answer different questions and get confused constantly in meetings.
Weekly is right for deal inspection on a short cycle, and fortnightly is usually enough when the average deal takes a quarter or more. The thing to protect is the gap between inspection and the quarterly number: if you only look properly at quarter end, you have a reporting ritual rather than a management system.
Keep the room small. Reps whose deals are being inspected, the manager, and nobody who is there to listen. An audience changes what people admit, and the entire value of the meeting is in what gets admitted. If other functions need the output, send them the output.
Whoever runs it should arrive having read the pipeline already. A manager reading the CRM for the first time in the meeting is the reason these sessions turn into narration. The hygiene that makes that possible is unglamorous and it is covered in CRM hygiene for small sales teams.
The output of a review is a short list of decisions, not a transcript. Deal, decision, owner, date. If that list does not exist by the end of the hour, the meeting did not happen in any sense that affects next week.
Recording and summarising the conversation is the one part of this that tooling genuinely helps with, as long as you are clear about what it is good for. Summaries capture what was said, which is useful; they do not decide what is true, which is the manager's job. The honest read on what these tools do and do not do is in rivl's note on AI meeting notes for teams, written for the same small team context.
A pipeline review cannot fix a pipeline that is too small. If your coverage is thin, inspection will tell you faster that you are going to miss, which is valuable, but the work is upstream. Work out the ratio you actually need first: pipeline coverage ratio is the companion to this, and how to build a B2B sales pipeline you can forecast is where the structure comes from.
We are also not going to quote a figure for how much a disciplined review improves forecast accuracy. The numbers circulating for this, claims that most forecast misses trace to poor reviews or that inspection cuts late stage losses by a fixed percentage, appear in vendor material without a published method or a control group. A number like that is worse than no number, because it becomes a target. What we can say is directional and mechanical: deals that move categories on evidence produce a forecast that is wrong less often than deals that move on sentiment.
Last caveat, and it is the one that decides whether any of this holds. A review only works if killing a deal is a safe thing for a rep to do. If the culture punishes a shrinking pipeline, people will keep dead deals alive in late stages, and you will have built an inspection process that reliably produces fiction. Fix that first, because no agenda survives it.
Send three things: your current committed number, how many of those deals have had a date move more than once, and the last unprompted thing each buyer did. We will come back with which deals we think are real, which are interested conversations, and what the honest forecast looks like.
Book a Free Consultation