# Which win-loss findings justify a message change, and how to test it

Decide which win-loss findings warrant a correction, a test or a hand-off, write wording you can substantiate, test the open question, and keep the evidence.

URL: https://www.hiresteve.ai/blog/win-loss-to-messaging
Updated: 2026-10-12

Start with one message and the buying decision it is meant to help, then decide what the reviewed findings warrant for that message: a correction, a test, or a hand-off to another owner. The output is a short change brief that pairs the proposed wording with its buyer evidence, the contrary evidence, the product fact behind it, the question a test must answer, and who decides. One verified counterexample can justify withdrawing an absolute claim. A recurring preference earns a test before it earns a place in the messaging.

## Start with the message you may need to change

This lesson begins after the interviews have been [analysed into reviewed findings](/blog/analyze-win-loss-interviews), each with the excerpts and interview or deal IDs behind it. If the batch includes CRM loss reasons that disagree with what buyers said, [reconcile them first](/blog/win-loss-vs-crm). [Win-loss without a research team](/blog/win-loss-analysis-without-a-research-team) shows where findings go; this page decides whether a finding deserves to go there.

Pick one message and the audience whose buying decision it serves: a pricing-page statement, a deck slide, a battlecard line. Have in hand the reviewed findings, the original excerpts with IDs and dates, a manifest of which segments, outcomes and questions the batch covered, the current approved wording, and the product facts behind it. Use redacted, permitted material. Anything missing stays missing; a polished synthesis cannot supply an outcome label the notes never recorded.

The change brief is the artifact. One per message, with these fields:

- Current wording, and the asset and audience it appears in.
- The observation, with the excerpt and interview or deal ID behind each supporting row.
- Contrary evidence: any buyer or deal in the batch that points the other way.
- The proposed explanation, marked as a hypothesis.
- Proposed wording, and the product fact that makes it true today.
- The question a test has to answer, and how you will observe it.
- Owner and decision: adopt, revise, hold or route, with the date.

## Check the finding's scope before you count anything

Two operations get confused here. Counting deals tells you how many separate buying decisions in this batch showed the finding. Corroborating sources tells you whether one account holds up. Outreach's guide to running a win-loss analysis advises [requiring at least two independent sources before acting on a pattern](https://www.outreach.ai/resources/blog/how-to-run-a-win-loss-analysis). That is advice about corroboration. It does not say a finding needs two buyers before a message may change.

Group the supporting rows by deal and segment. Several excerpts from one deal are one buying decision, though two people describing the same deal can still corroborate a detail. If every interview was a loss, or every buyer came from one segment, write that at the top of the brief, because a finding cannot recur in a segment you interviewed once. Counts describe the supplied batch, never how common something is among buyers in general.

Ask Claude to do this grouping in a chat inside [the Project you keep for product marketing](/blog/claude-project-setup). Anthropic's Projects page says you can [upload documents, provide context and have focused chats](https://support.claude.com/en/articles/9517075-what-are-projects) within a Project, and its management page says context is [shared across a Project's chats only through the knowledge you add](https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects). Supply the current findings sheet, the manifest and the approved wording as files; an earlier chat is not an evidence pack. Ask Claude to list what it received and what is missing, then to attach the actual excerpts to each finding, keep buyer statements separate from seller interpretation, and group by deal and segment. Then open the originals behind each candidate yourself.

## Decide whether it is a correction, a test, or someone else's problem

**A verified false claim does not wait for a second interview**

If one buyer shows that an absolute statement in your messaging is wrong, and you can confirm that against the product, withdraw or narrow the statement now. The correction depends on the product check rather than on a second buyer. Send an urgent single-claim fix through the [battlecard feedback process](/blog/battlecard-feedback); this page handles the batch.

Every other finding gets one of three dispositions. A plausible issue raised by several separate deals, or by one deal in a way you expect to recur, becomes a test candidate: the brief proposes wording and a test, and nothing ships until the test has run. Thin or contradictory evidence stays open, with what would resolve it written in the row. A finding that describes a product gap, a procurement constraint or a sales-process failure goes to the owner of that problem with its scope attached, because rewriting a sentence will not fix it.

That last group includes losses that look irrelevant. A parent company that mandated another vendor, or a budget that disappeared mid-cycle, is a finding about qualification or segment fit, and several in one segment may change how sales qualifies. Retain each with its scope and route it. There is no minimum interview count for any disposition and no automatic discard of minority evidence. Each row has to say why wording is the intervention, or why it is not.

## Write the sentence you can substantiate, and keep the hypothesis separate

A finding has three parts that belong in three places. The observation is what buyers said. The explanation is your guess at why. The wording is what a rep or a page will say next. Writing them as one sentence turns a guess into a claim.

An illustrative example, invented for this lesson:

| Part | Text | Where it lives |
| --- | --- | --- |
| Observation | Buyers in two separate deals (IDs 14 and 22) said they could not tell from the website which plan fit them. | Findings sheet, with excerpts |
| Explanation | The pricing page describes plans by feature list, and neither buyer could map their usage onto it. | Brief, marked as a hypothesis |
| Proposed wording | "Plans are sized by monthly usage. The first call confirms which band you are in." | Brief, only if plans really are sized that way |
| Test question | Do buyers who hear this sentence identify their band correctly, and does anyone still pick the wrong plan afterwards? | Brief |

The proposed wording promises only something the product already does. It does not say the walkthrough makes buyers choose correctly, because two buyers reporting confusion cannot show that. If the factual proof is missing, request it or choose a clarification that is already true. A sentence a future buyer could contradict is a good hypothesis; it becomes a message once the product fact is confirmed and the test has run.

Compare the draft with the approved messaging before anything else. Wording that contradicts an approved position raises a [positioning question](/blog/positioning-brief) and goes through that decision before any card changes. Wording that survives the check takes its final shape in the [messaging template](/blog/messaging-template).

## Test the uncertainty, and record what happened

The test answers the question in the brief, nothing larger. A person arranges limited use of the approved factual wording in conversations where the issue arises, and records three things: whether the wording was used at all, what the buyer understood by it, and any objection or counterexample that came up. Set the review point by the actual opportunities to observe that situation. When those opportunities are rare, a short fixed window may end without a usable observation.

Keep three results distinct. "Not used" means the reps did not say it, which is its own finding about the wording or the training. "Used but misunderstood" means the sentence needs work. "Understood but did not address the real problem" means the explanation in the brief was wrong, and the finding goes back to open. A small test with a few reps can show whether wording is usable and understood. It cannot show that the wording changed win rate; a handful of calls cannot carry a causal claim.

Feed the permitted observations back into the Project chat. Ask Claude to compare the supplied usage notes and outcome labels with the hypothesis, then check every flagged mismatch against its source ID and original record. Calls nobody recorded remain absent, so the observations have to arrive as files. Adoption stays a human decision.

## Keep the evidence when you keep or reject the change

Record the decision (adopted, revised, held or routed), the owner, the assets affected, the evidence dates and the next check, whichever way it went. Update the approved messaging file in Project knowledge explicitly, and keep the superseded wording and the source excerpts beside it, so a future reader can reconstruct why a sentence changed.

Later batches will produce findings that resemble held ones. A resemblance prompts a review rather than an automatic promotion. Before treating two observations as the same finding, check that the product version, the segment, the question wording and the period are comparable. A complaint about plan selection before a pricing-page rewrite and one after it are two findings.

## Where Steve Win/Loss helps with the inputs

If recurring buyer-evidence collection is the step your team misses, schedule outreach, keep one current evidence file and hold a short review with the message owner.

Steve Win/Loss sends a CRM-triggered buyer survey when a HubSpot deal or Salesforce opportunity closes, after the delay you choose. A survey can be triggered automatically when a deal closes, or sent by link manually.

Click Generate Summary when new responses arrive; each summary is built from the responses available at that moment. The summary includes an overview, why you win, why you lose, competitor insights and cross-question patterns. Steve detects themes with AI, so check them against the buyers' own responses.

Steve Win/Loss surfaces buyer themes for your team to review against battle cards and objection answers; your team still chooses, tests and records the wording. The tradeoff is setup plus ongoing report verification. If recurring collection is the gap, see [Steve Win/Loss](/products/win-loss). If you already collect enough evidence, the brief and its decision record fit in a shared document.

## Frequently asked

### How many buyers have to raise a finding before I change a message?

There is no validated minimum. One buyer who exposes a false absolute claim, confirmed against the product, justifies withdrawing it. A recurring preference becomes a test candidate. Thin or contradictory evidence stays open. Outreach's advice to require two independent sources concerns corroborating one account; it sets no count of buyers.

### What do I do with a finding only one buyer raised?

Keep it with its scope and decide what kind of response it warrants. A verified factual error gets corrected. A plausible issue becomes a test candidate. A product gap, procurement constraint or sales-process problem goes to that owner. Nothing is discarded for being a single mention.

### Can Claude sort the findings for me?

For this workflow, ask Claude to use material supplied in the current chat or explicitly added to Project knowledge. Ask it to attach excerpts, group findings, and draft the brief. Missing records and outcome labels stay missing. Check each candidate against its source ID and original excerpt yourself.

### How do I test a new message before it goes on every card?

Arrange limited use of the approved factual wording in conversations where the issue arises. Record whether sellers used it, what buyers understood and any counterexample. A small test can check usability and comprehension. It cannot show a change in win rate.

### Does Steve turn win-loss findings into messaging?

Steve detects themes across buyer responses and shows them beside the buyers' own words. Your team reviews them and decides what changes in positioning, objection answers and battle cards, then tests and records the wording.