Win-Loss Interviews That Change the Message and the ICP

The method is yours: read the deals, hold the calls, label the notes and decide. Free tools give a first draft at three points, and you check each one. Each step feeds the next by hand, because no tool reads the output of another. A spreadsheet with one row per deal counts just as well. Keep one anyway, for the one reason that decided each deal. What the tools save is the blank page.

My suggested rule: two wins and no loss makes a lead, never a change. A lead becomes a change after four checks. Merge labels that mean the same and count again. Test rival explanations on every deal. Check "no loss" against who you interviewed and against public facts. Then try the line on the person who signs, and measure it in a trial.

Losses can propose too: the same pattern in two or more losses and no win can move an ICP field or a sales step. A pain in wins and losses alike is why buyers take the call, so it stays as the demand line. Only what separates wins from losses can be a differentiator.

Theme labels: GOV.UK, Analyse a research session | Rival explanations, missing buyers and exceptions: HM Treasury, Quality in qualitative evaluation | Sources opened 1 November 2026

Who invites: not the rep. Clozd says third-party interviewers "reduce social desirability bias and raise candor", and it sells that service. With no budget, use someone the buyer never negotiated with: a product marketer, or a founder who was not on the deal. If the founder sold every deal, use an outsider for the losses, or accept the bias and say so.

Ask within six weeks of the decision, since Clozd says recall "fades after ~6 weeks". Ask for 30 minutes. Offer a modest thank-you where the buyer's policy allows it. Ask before recording. Say that notes stay internal and nothing goes public without approval.

A suggested invite: "We are learning from our recent evaluation with your team, not reopening it. Could I have 30 minutes on how the decision was made? No sales follow-up will come from this."

Expect few yeses. Octave, also a vendor, says "Typical acceptance rates range from 20-40%" for lost buyers: two or three calls from eight asks. Clozd suggests 15 to 25 interviews per segment per quarter. My suggested rule: with fewer than six calls, or fewer than two wins and two losses, keep a list of leads, change no line and pool the next quarter's calls. Count no-decision deals as their own group, since no competitor won them.

Do not plant your hypothesis: ask who cared, not whether sales cared. Log every decline. The deals you cannot reach may be the ones that break your story.

Step 1. Read the quarter's deals side by side

About this example. SpotDraft is a real company. SpotDraft has not reviewed this guide. Facts about SpotDraft come from public sources, linked alongside with the date I checked them. Everything else here is illustrative: the deals, wins and losses, buyers, customer types, reviews, results and every number. None is SpotDraft's data, customers or capability. This is a worked example, not a client result or SpotDraft's real strategy.

What is public: SpotDraft describes itself as "AI-native CLM for in-house legal", for teams at growth-stage and mid-market companies. It raised an $8 million Series B extension in January 2026 to expand its enterprise presence across the Americas, EMEA and India and "accelerate adoption of on-device AI for secure contract lifecycle management".

If you ran go-to-market at a company like SpotDraft, a first win-loss round in one market gives you a baseline. This exercise is that round: eight illustrative deals from one quarter, each invite sent within six weeks of the decision.

Input to the deal portfolio run: analysis_type deal_portfolio, deal_outcome mixed, your_sector saas, the lines alongside in multiple_deals, and your_solution "SpotDraft, an AI-native contract lifecycle management platform for in-house legal teams at growth-stage and mid-market companies".

What came back: "You listed 8 deals: 3 won, 4 lost, 1 no decision". How to read it: the tool counts and sorts; the pattern is yours to find. In two of the three illustrative wins a sales leader pushed for the purchase. No loss line names one, but three of the four are silent, so this is a lead, not a fact. Two lines say price, and that stays open until a buyer speaks.

Step 2. Debrief the rep, then interview the buyer

Input to the single-deal run for deal 4, all illustrative: analysis_type single_deal, deal_outcome lost, the same your_solution and your_sector, deal_value 38000, sales_cycle_days 84, loss_reason "price", competitor_won "the contract add-on of their existing e-signature vendor", and the two text fields alongside.

What came back: the tool classed the winner as "an incumbent or legacy setup" that a buyer keeps "when it is already paid for, integrated and approved", and returned seven questions.

How to read it: the questions are a standard set filled from your deal and worded to the seller, so they are a first draft for the rep debrief. The buyer call uses the script above, held by someone else. The tool lists "VP Sales (not met)" and leaves the meaning to you, so let the buyer raise it unprompted. Price is still open: pricing came once, in the final week. The script's price question tests it.

Do this for all eight deals. Assume, for illustration, that six of eight buyers agree: the three wins and the logistics, edtech and cybersecurity losses. That is well above what to plan for; it keeps the exercise full. Assume the martech buyer and the proptech lawyer decline. Those deals stay in the portfolio read and out of the count, because a rep's note is not a buyer's words. One more call could end any "no loss" below, so no decision here rests on one alone.

Step 3. Put the notes in one wording, then count wins and losses apart

Input to the interview synthesizer: company "SpotDraft", product_category "AI-native contract lifecycle management for in-house legal teams", and the six illustrative interviews alongside, with key_quotes left out.

Before you run it, give each idea one label. The tool counts matching wording, not meaning, so merge variants by hand and keep buyers' own words in your sheet, with the one reason that decided each deal. A pain someone mentioned may have played no part, and the tool has no field for that.

What came back from all six: top pain "no view of where a contract sits", "in 4 of 6 interviews (67%)", and the same line as the pain message. Those mentions split two wins and two losses, so the line does not explain why anyone signed. I ran it twice more, wins only and losses only.

On wins only, "sales contracts stuck in email threads" and "no view of where a contract sits" tie at 2 of 3, and a tie is yours to judge. The top trigger is clearer: "new VP Sales hired" in 2 of 3 wins and no loss, while "contract volume doubled" leads the losses.

Every interview is read as a customer, so treat the losses-only "Add to ICP" list as what not to add. "Champion Roles" counts who you interviewed, so take champions from your deal lines.

Step 4. Test each lead, then decide

This step is yours, not a tool's, and every result here comes from illustrative inputs, not SpotDraft's positioning or ICP. Why so strict? With three wins and three losses, a label unrelated to the outcome and carried by one account in three clears "two wins, no loss" about 8% of the time. Across this exercise's ten labels, the chance that at least one clears it by luck is about 55% (my sum, treating labels as independent).

First lead: the email line, in two wins and no loss. It fails the merge test. A contract in an email thread is a contract nobody can see. Replace that label with the no-view label, once per interview, and rerun: 5 of 6, with 3 of 3 wins and 2 of 3 losses. The difference came from the labels, not from buyers. So "no view of where a contract sits" stays as the demand line.

Second lead: a sales leader. Deals 1 and 3 share two other traits, so the rule cannot choose between the stories alongside. "No loss" means only that three people did not say it, and no revenue leader spoke in any loss. So check all eight accounts for a new sales leader in public hiring news, and hear one revenue leader from a lost deal. The sponsor belongs in qualification: nobody can see it on an account list, and it would have screened out Deal 2.

Losses propose the ICP change. Both deals over 1,000 staff were lost, and the sole lawyer stalled. So test two fields you can filter on: 300 to 1,000 staff, and at least two lawyers. Larger accounts become a separate track, since the public plan is to expand the enterprise presence.

The message keeps the demand line and adds the sales cost of slow contracts as the sponsor's case. That could tell a legal buyer that legal is the bottleneck, and SpotDraft's home page promises speed "with legal in charge". So try it on two legal signers first, then with two reps for a month.

Two threads need an answer. The add-on and the suite: sales prepares a reply to each. The AI security review: deals 4 and 7 met one, three weeks and 70 days. That could be the buyer's calendar or a gap in the story, so ask the reviewer what would have shortened it and whether they heard the on-device aim.

This is the point where I would want a second pair of eyes: buyer findings conflict before you change the message or the ICP. Bring me your separate win-loss counts, declines, rival explanations and proposed changed lines, and we will work it through in 45 minutes.

Step 5. Ask the right customer for proof

If the sponsor's case survives its test, it needs a revenue leader's measured result, but Deal 3 has none yet. So the request goes to the CRO at an illustrative earlier customer, won seven months ago. The input to testimonial capture in CRAFT Content is alongside, with placeholders.

What came back: a request email, questions for the role and quote starters, with a note that nothing is a quote until the customer approves it. Treat it as a first draft: fill the placeholders from the customer's own report, rewrite the opening, and cut questions that do not fit a customer seven months in.

1. Taking the CRM reason as the finding.
2. Counting wins and losses together.
3. Counting raw notes. A label you drew can create a difference no buyer made.
4. Letting the rep run the interview, or reading a buyer questions worded to the seller.
5. Ending with a readout that changes no line, or changing a line on two wins with no test.
6. Quoting an interview as a testimonial without the customer's approval.

• From the Buyer Who Champions You to a Message Sales Will Use: Positioning With IMPACT, CRAFT Content and Revenue Enablement
• How to Turn One Published Customer Result Into a Case Study, a LinkedIn Post and a Newsletter Without Changing the Number
• All 24 guides