Win/Loss Pattern Synthesis

The required picklist in the CRM is not a debrief. It is the sentence the seller needed to file before the quarter closed.

Partners choose Price, Relationship, Incumbent, or Timing because the field is mandatory. Buyers say price when they mean they did not believe the team, or procurement needed a reason that would survive an audit. Feed those codes to a model and it will "discover" that you lose on price and win on relationships. That is the picklist talking back.

The useful job is clustering themes from interviews and artifacts: what a buyer on the evaluation said, what you submitted, how the deal was scored if you can get it, who beat you, and how that differs by segment and deal type. Dedicated win/loss firms and platforms such as Clozd exist to run those interviews at arm's length. Conversation intelligence such as Gong can show what happened on your calls. Survey tools such as Qualtrics can collect a structured follow-up. None of them are a dashboard of Price vs Relationship from a required field. If that dashboard is the deliverable, skip the model.

A Closed-Won or Closed-Lost stamp tells you the deal ended. It does not tell you why.

Capture or BD ops should assemble a pack per deal while people still remember it:

• Internal debrief notes written within a week, with who was in the room and what they disagreed about.
• A buyer-side interview with someone who sat on the evaluation, not only the coach who liked you. Use a third party when the account team cannot hear a straight answer.
• The submitted proposal, plus the scoring sheet or buyer debrief email when you can get it.
• Competitor identity if known, and whether you were incumbent, challenger, or a courtesy bid.
• Labels a partner would defend: new logo vs existing, commercial vs public sector, competitive RFP vs sole-source, practice, industry, fee band.

Interviews beat coded reasons. A forty-minute call will contradict the CRM field more often than it confirms it. Surveys are cheaper and more polite. They under-sample people who have moved on and invite the same "price" answer. Use a survey after an interview program exists, not as the corpus.

Gong-class recordings are useful for the sell-side story: where the team hedged on staffing, skipped a diagnostic, or never asked for evaluation criteria. They are not a substitute for the buyer interview. If you cannot get buyer interviews on losses, say so. Synthesizing only internal notes will encode the account team's self-defense.

A finding that is true across the whole firm is usually true of nothing you can change on Tuesday.

Public-sector framework bids, mid-market commercial RFPs, and partner-led sole-source expansions do not share a win condition. Mix them and you get "we need to be more commercial." Slice first: industry, buyer type, new logo vs incumbent, competitive vs not, and the offer (diagnostic, implementation, retained advisory). Cluster inside a slice.

The model will merge "generic workplan" on manufacturing RFPs with "thin security appendix" on a bank RFP into one "proposal quality" theme. Keep each weakness attached to the section and the segment where it showed up.

Only then should a slice-level theme become a question in a bid/no-bid opportunity classifier. A firm-wide "we lose on price" flag will no-bid work you should take and still send you into the work you keep losing.

Four Closed-Lost rows coded as Price can still be four different problems. The following is illustrative, not a measured result.

A 250-person boutique runs a supply-chain operations practice. Across two quarters they lose four mid-market manufacturer RFPs, all Closed-Lost with reason Price. A partner wants to cut the rate card. Capture pulls the four submitted proposals, the CRM notes, and runs buyer interviews, the same job a Clozd-class program or an in-house capture lead with a script would run.

Deal one: the buyer said the workplan could have been written for any plant. They awarded a firm that priced a two-week on-site diagnostic inside the proposal. This boutique had treated that diagnostic as unpaid BD. "Price" was the polite close.

Deal two: they were the incumbent on a smaller warehouse project. The operations VP had been trying to exit for a year. The interview named trust, not fee. The engagement partner still filed Price.

Deal three: a real rate-card gap against a lower-cost specialist on a bounded inventory-optimization scope. The proposal was specific. The buyer said they would have paid a modest premium for a named partner they already knew, and they did not know anyone on this team.

Deal four: a courtesy bid. The incumbent was never going to be displaced. Internal notes said as much. Nobody had marked it a no-bid.

A model run on the four CRM fields outputs one theme: price. Acting on it would cut rates on the next new-logo manufacturing RFP, still skip the diagnostic, still staff unknown names, and still chase courtesy bids. Interviews plus artifacts split into different actions: productize a priced diagnostic for new-logo plant work, refuse to bid unless the delivery partner is in the room, and tighten the bid/no-bid screen on courtesy incumbents.

Those four files are still hypotheses. Hold "generic workplan on new-logo manufacturing RFPs" until it shows up in more interviews and in the submitted documents: no plant-specific constraint, no named diagnostic, credentials from another industry. Until then it is a note on four deals, not a firm finding.

Four similar notes will always look like a pattern because people and models are good at completing a sentence.

Set a floor before you publish a theme: enough deals in that slice, plus agreement between the buyer interview and something in the artifact (the proposal, the scoring comments, the competitive alternative). If only the account team says "price" and the proposal is vague, believe the proposal. If one loud partner insists "relationship," treat it as that partner's theory until another slice repeats it.

Recency bias shows up fast. Last month's embarrassing loss will dominate the cluster if you do not weight by slice and by whether the bid was real. Wins need the same discipline. "We win on senior time" is often the story you tell after a sole-source. Check whether that theme appears in competitive RFPs in the same segment.

Do not publish a theme that reads as a named partner's failure. Frame it as a pursuit or proposal problem the next team can change: missing diagnostic, unnamed staff, wrong proof point, no map to the evaluation criteria. Findings that sound like blame get quietly ignored, which is how you keep losing the same way.

A lessons learned synthesizer on delivery is a different corpus. Do not mix delivery regret with bid-stage themes unless the buyer actually cited delivery on the incumbent work.

A theme that does not change a screen, a draft, or a red-team check is a slide.

When a theme holds in a slice:

• Pursuit screen. Add it to the bid/no-bid opportunity classifier as a question a capture lead can answer this week: will they fund a diagnostic, will the delivery partner be in the room, is this a courtesy bid.
• Drafting. Tell the RAG proposal draft generator which prior sections actually scored in that segment, and which "winning" files were losses nobody relabeled.
• Red team. Put recurring evaluator objections into the proposal red-team reviewer so the next draft is scored against what this buyer class already dinged you for.
• Proof. Only turn a win into a reusable case study when the interview said that proof point mattered. A case the team loves and the buyer never mentioned is not evidence.

Judge the program on whether the next bids in that slice look different, and whether later interviews still name the same weakness. If the dashboard now says "solution fit" instead of "price" and the proposals are unchanged, you automated the picklist.

Re-run on a schedule, by slice, with new interviews. Do not re-cluster last year's CRM export and call it an update.