A rep loses a deal and types "price" into the CRM. It's quick, it's plausible, and it's usually wrong, or at least incomplete. The buyer didn't leave because of your price; they left because they never quite believed you'd deliver, and "too expensive" was the polite way to say so. Win-loss analysis exists because the story the seller tells about a deal and the story the buyer would tell are rarely the same story, and only one of them can improve your win rate.

The quick version

  • Win-loss analysis is the practice of systematically interviewing buyers, both the ones who chose you and the ones who didn't, to learn why deals were really won or lost.
  • The source you can't skip is the buyer. Sales reps' loss reasons are educated guesses; CRM fields are the guess, abbreviated. The only person who knows why a buyer bought is the buyer.
  • Your biggest competitor often isn't a rival, it's "no decision." Research on millions of B2B sales calls found 40–60% of qualified deals stall out with no purchase at all.
  • The honest caveat: buyers misremember and rationalise too. Treat each interview as evidence to triangulate, not a verdict, patterns across many conversations beat any single quote.

The idea in depth

Why the loss reason in your CRM can't be trusted

Start with the uncomfortable premise. When a salesperson records why a deal was lost, they are reporting a decision they didn't make, weren't in the room for, and have every incentive to read charitably. The reliable way to find out why buyers buy is to ask the buyers, which is the whole proposition of win-loss analysis. Steve W. Martin, who has run this kind of research and written it up repeatedly for Harvard Business Review, makes the point bluntly: there's a strong tendency to assume a deal was lost because the product was inferior, yet when you actually interview decision-makers, they tend to rank the competing products' feature sets as roughly equivalent ("6 Reasons Salespeople Win or Lose a Sale," HBR, 2017). If the features are a wash, the deal turned on something else, trust, the buying experience, perceived risk, and that "something else" is invisible from your side of the table.

The fix is to separate two record-keeping jobs you've probably merged. Keep the rep's quick CRM tag for pipeline hygiene, but stop treating it as the reason. Build a second, slower channel, a real conversation with the buyer, conducted by someone who isn't the rep, and let that be the source of truth about why. The rep's note tells you a deal closed; the interview tells you what to change.

Interview the wins and the losses, and the deals that went nowhere

The instinct is to autopsy the losses, because losses sting. That's half a programme. Win interviews are where you discover the things you do well by accident and could do on purpose, the offhand reassurance that tipped a nervous buyer, the reference call that closed it. Without the wins you optimise for the absence of complaints rather than the presence of what actually persuades.

And there is a third category most teams never interview at all: the deals that ended in nothing. In their book The JOLT Effect (2022), Matthew Dixon and Ted McKenna analysed more than 2.5 million recorded B2B sales conversations and found that somewhere between 40% and 60% of qualified deals end not in a loss to a competitor but in no decision, the buyer simply doesn't buy anything. More striking, their research attributes the majority of those stalls, around 56%, to customer indecision and fear of getting the purchase wrong, rather than a genuine preference for the status quo. If you only interview won-and-lost deals, you're blind to the largest competitor you have. Add a "no-decision" track to your interview list, then ask it a different question: not "why them?" but "what made buying at all feel too risky?"

flowchart TB
  D("A qualified deal closes, won, lost, or no decision")
  R("Rep logs a quick reason in the CRM")
  I("Independent interviewer talks to the buyer, 1–2 weeks later")
  P("Patterns across many interviews")
  A("Change the product, message, or sales process")
  D --> R
  D --> I
  I --> P --> A
Two channels from one deal: the fast CRM tag for hygiene, the slow buyer interview for truth. Leaders Loop

Who asks, and how soon, decides what you learn

Two design choices quietly determine whether a programme yields signal or flattery. The first is who conducts the interview. When the salesperson on the deal runs the debrief, the buyer softens the truth to spare their feelings, and the interviewer hears what confirms their own account. The practitioner consensus, see the Pragmatic Institute's win-loss best practices, is to put a neutral party in the chair: a product or research colleague, or an outside firm, so the buyer can be candid and the analysis stays unbiased. The implication is simple and slightly uncomfortable: take the rep out of the interview, even though they want to be in it.

The second choice is when. Memory of a buying process decays fast, and it decays in a particular direction, toward the tidy, self-justifying narrative the buyer has since told their own boss. Run the interview within a week or two of the decision, while the texture is still there. Make the trigger automatic: when a deal hits closed-won, closed-lost, or no-decision, an interview request goes out the same week, before the story hardens.

Where this breaks down. Win-loss analysis inherits the limits of all self-report. Buyers don't have perfect access to their own reasons; they rationalise, they flatter, they blame price because it's socially easy. A single interview can send you confidently in the wrong direction. The discipline that saves you is triangulation: look for what repeats across ten or twenty interviews, cross-check it against your CRM data and the rep's own debrief, and trust the pattern over the anecdote. If three buyers in a row, unprompted, describe the same moment of doubt, that's a finding. One vivid quote is a hypothesis, not a mandate.

A worked example

Picture a mid-market software company, call it Meridian, with a win rate stuck around an illustrative 22% and a sales team certain the problem is pricing. (Illustrative figures throughout; the shape is realistic, the numbers are invented to show the method.) Leadership is one discount away from cutting list price across the board. Before they do, they run twenty win-loss interviews, eight wins, eight competitive losses, four no-decisions, using a product manager, not the deal's rep, as the interviewer, each call inside two weeks of the decision.

The CRM had labelled fifteen of those twenty deals "price." The interviews tell a different story. Among the competitive losses, buyers consistently say Meridian's demo felt generic, the rep showed features, the winning vendor showed their workflow, and that a slow, vague answer on data security late in the process quietly killed their confidence. Price came up, but as a justification, not a cause: "I couldn't get comfortable they'd handle our migration, so the premium wasn't worth it." Among the four no-decisions, a clear pattern emerges, the buying committee couldn't agree internally and, fearing a costly mistake, chose to do nothing. That's the JOLT pattern in miniature: not lost to a rival, lost to indecision.

What does Meridian actually change? Not the price. They rebuild the demo around the buyer's workflow, arm reps with a crisp two-minute security answer and a reference customer, and add a "mutual close plan" that helps a divided committee reach a decision they won't regret. They've turned three buyer quotes into three concrete changes, one to the product story, one to the sales process, one to the deal mechanics. A quarter later the win rate moves, and the discount they nearly gave away is still in the bank. Notice what the analysis did: it didn't hand them an answer, it replaced a confident guess with evidence, and the evidence pointed somewhere they'd never have looked.

flowchart LR
  C(["CRM says: 'Lost on price' (15 of 20)"])
  B(["Buyers actually say: generic demo, shaky security answer, committee couldn't decide"])
  M(["Fixes: workflow-led demo · crisp security proof · mutual close plan"])
  C -.the guess.-> B
  B --the evidence--> M
Meridian's gap between the logged reason and the real one, and the three changes the interviews actually justified. Figures illustrative. Leaders Loop

Frequently asked questions

Isn't this just asking the sales team why they lost?

No, and that substitution is the most common way the practice fails. The salesperson is a witness to the deal, not to the buyer's mind; they weren't in the room when the decision got made, and they have a natural stake in the explanation. A debrief with your own rep is a useful input, but it's a hypothesis about the buyer's reasons. Win-loss analysis means going to the buyer themselves, ideally through a neutral interviewer, and treating the rep's account as one source to cross-check rather than the answer.

How many interviews do I need before I can trust the findings?

Enough to see a pattern repeat, which is less than people fear. Because you're looking for recurring themes rather than a statistically representative sample, ten to twenty good interviews across wins, losses and no-decisions usually surfaces the two or three issues that matter. The test isn't sample size, it's saturation: when new interviews stop telling you anything new, you have your signal. One interview is a story; the fifth time you hear the same story, it's a finding.

Will buyers, especially the ones who rejected us, actually talk to us?

More than you'd expect, particularly the losses. Buyers who chose a competitor are often willing to explain why, especially when the request comes from someone other than the rep who was just selling to them, the conversation is short, and it's framed as "help us get better" rather than "let us win you back." A neutral interviewer and a no-pitch promise do a lot of the work. Wins are easy; the no-decisions are the hardest to reach, because there's no relationship to lean on, which is exactly why a fast, low-friction request matters.

How is win-loss analysis different from NPS or a satisfaction survey?

Different question, different moment. Satisfaction and loyalty metrics like NPS measure how existing customers feel about you over time; win-loss analysis examines a specific buying decision at the moment it's made, including by people who never became customers. NPS tells you whether to worry; win-loss tells you what to change about how you sell and what you sell. They're complementary, not interchangeable.

We're a small company with no research team. Can we still do this?

Yes, and the constraint can help. You don't need software or an agency to start, you need a product manager or founder (anyone who isn't the deal's rep) to make five calls this quarter with a short, consistent question set, take honest notes, and look for what repeats. The method scales down cleanly because its value is in the listening, not the tooling. Start with five losses and one question you genuinely don't know the answer to.

Related in the Toolkit

Where to go next