Win/loss analysis has a selection bias problem
The process looks rigorous, which is why the flaw is easy to miss. You take the deals that closed, interview the buyers, code the reasons, feed it back to sales and product. The output is a clean picture of why you win and why you lose to named competitors.
But every person in that sample crossed the decision threshold. The people who never crossed it are largely absent from the data. The buyer who evaluated you for four months, went quiet, and bought nothing from anyone did not choose a competitor, so there is no competitor to code, and did not choose you, so there is no deal to review. In most CRMs they become a closed-lost record with a reason code like "budget" or "timing" — entered by the rep who was ghosted, not by the buyer who ghosted them.
This is selection bias in the ordinary research sense: the sampling method excludes the outcome category you most need to understand, and it excludes it systematically rather than at random. The result is a picture of the deciding minority that gets read as a picture of the market.
This is why the picture stays stable no matter how many win/loss cycles you run. You are refining your understanding of the 40 percent of the market that decides, while the majority that does not stays exactly as opaque as it was.
Reason codes describe the last conversation, not the first cause
The gap gets papered over by the one artefact that does exist for these deals: the closed-lost reason code. Budget. Timing. Executive priorities. Internal resources. Not a priority this cycle.
Every one of those describes how the deal ended rather than why it stalled. They are the polite, defensible answers a buyer gives when a conversation needs to conclude, recorded by the person on the other end of the silence. Underneath, the same handful of causes recur — the risk of choosing wrong, an internal group that never aligned, a job that turned out to matter less than it seemed, an anxiety nobody addressed because nobody heard it.
Coding those as budget and then reporting on budget is how a demand problem becomes a pricing conversation.
Two causes, opposite fixes
Matthew Dixon and Ted McKenna analysed 2.5 million recorded sales conversations for The JOLT Effect and found something that should change how founders read a stalled pipeline. No-decision losses are not one phenomenon. As the figure above shows, they split roughly in half — and the smaller share is the one most sales training is built around.
About 44 percent never accepted that a problem existed. The current way is good enough, the switching cost is real, nothing forced the issue. Habit is doing the work. The other 56 percent accepted the need to change and froze anyway. They wanted to move; the fear of choosing wrong outweighed the cost of staying put. Anxiety is doing the work.
The two look identical from outside — both end in silence — and they require opposite treatments.
| Buyer state | Root cause | Wrong response | Right response |
|---|---|---|---|
| Status quo | No urgency; the current way is good enough | More proof the product works | Build dissatisfaction with staying |
| Indecision | Fear of choosing wrong | More urgency, more options, a discount | Reduce perceived risk of the choice |
Which means the default reaction to a stalling pipeline — more ROI slides, another case study, executive pressure, a procurement discount — treats indecision as though it were indifference. For the majority of these deals it is not merely ineffective; it makes the freeze worse.
The part the research cannot tell you
Dixon and McKenna prove the two causes exist and behave differently across the market. What that research cannot tell you is which one is killing your deals, in your category, or what the anxiety is actually about when it is your product on the table.
That is not a gap in their work. It is a limit of the method. Conversation analysis observes what buyers say to sellers; it cannot observe what buyers say to themselves, or to colleagues after the call ends — which is where the decision not to act is actually made. People do not tell a rep they are worried about looking foolish in front of their VP. They say they need to circle back next quarter.
So the split is real and the diagnosis is still missing. You know 56 percent is probably fear. You do not know what they were afraid of.
What to do instead: interview the people who didn't buy
The instrument for this is a no-decision interview — a jobs-to-be-done conversation with someone who had the struggle, wanted the progress, evaluated real options, and hired nothing. Christensen’s framing is useful here precisely because it accommodates that outcome: customers hire and fire products, and choosing to hire nothing is a decision with causes, not an absence of one. It is an entire population with demand in it, and almost nobody studies them.
It is also the hardest of the six interview types to run well. People rarely volunteer that they did nothing out of inertia or fear, so the interview has to make that admission feel normal before it can get anywhere.
Run properly it separates the two causes directly. Habit questions surface what made staying acceptable: what they were already using, what they would have had to give up, why nothing had ever been the right fit. Anxiety questions surface what made moving feel risky: what concerned them about the alternatives, what they wanted to avoid at all costs, what they were worried would happen. The answers arrive in the buyer's own words, which is also the raw material for the messaging that would have addressed it.
Ten to twelve of these will tell you more about your stalled pipeline than another year of closed-lost reason codes. They are also the least-run interview in B2B, and the reasons are practical rather than mysterious: the list is harder to build, response rates are lower, and the conversations are less comfortable than talking to customers who chose you. That is precisely why the learning is still sitting there.
The complete no-decision interview guide
Twenty-three questions across four phases, with what each one is designed to surface and which force of demand it belongs to. Free, no email required.
Open the guide →Research cited: Matthew Dixon and Ted McKenna, The JOLT Effect: How High Performers Overcome Customer Indecision (2022), based on an analysis of 2.5 million recorded sales conversations. The 44/56 split and the 40–60 percent range are theirs; the selection-bias argument, the reason-code critique and the interview method are ours. Christensen’s hiring and firing framing is from Competing Against Luck (2016).