All the best messaging in the world starts with listening. Not to your product team’s assumptions about what buyers want. Not to the marketing copy your competitors use. To your actual customers, in their actual words, describing their actual problems.
Customer research is the discipline of capturing that signal and turning it into inputs that sharpen the narrative, the ICP, the persona stack, and ultimately the messaging. It’s not glamorous. It also compounds faster than almost any other PMM investment.
Win/Loss Analysis
Win/loss is the single most underused source of PMM gold in most companies. Every deal that closes and every deal that doesn’t is a data point. Most companies capture neither systematically.
What you’re trying to learn: Why did we win? Why did we lose? What were the real reasons, not the CRM notes? Who was the deciding voice? What almost made the deal go the other way?
How to run it: Conduct short (20-30 minute) interviews with contacts from recently closed and recently lost deals. Within two weeks of the outcome if possible, while the memory is fresh. If you can’t get the lost customer directly, debrief the sales rep who ran the deal, with the explicit goal of separating what they think happened from what the customer actually said.
A few questions that consistently surface useful signal:
- “What made you start looking for a solution in the first place?”
- “What were the two or three things that mattered most to you in the decision?”
- “Who else were you seriously evaluating, and what made the difference?”
- “Was there anything that almost made you go the other way?”
What you’re listening for: the buyer’s language (not your language), the real buying trigger (often different from what brought them into your CRM), and the moments of doubt that you either overcame or didn’t.
Customer Interviews
Win/loss is transactional. Customer interviews are longitudinal. You’re trying to understand the texture of the job your customer is trying to do, the frustrations they carry, the language they use, and the definition of success they’re working toward.
Who to talk to: Your most successful customers and your most recently churned ones. Both teach you different things.
How often: Regularly, not just when you’re building something new. PMMs who only do customer research at launch time are flying blind between launches.
What to do with it: Synthesise the themes. What language do customers use repeatedly to describe the problem? What outcomes do they care about that your current messaging doesn’t address? What objections come up that your sales enablement doesn’t answer?
The candid moment after a customer call, the thing someone says after you’ve said goodbye but before they hang up, is often the most valuable data point in the whole conversation. Create space for it. Ask “is there anything else you’d want us to know?” and then go quiet.
Turning Research Into Inputs
Raw interviews aren’t messaging. Synthesis is the step most people skip.
After a round of research, pull all your notes and look for:
- Recurring phrases: the exact words buyers use to describe the problem. These belong in your messaging verbatim, not paraphrased.
- Unexpected buying triggers: the thing that made them start looking. Often different from what you assumed.
- Unstated objections: the concerns buyers had that they didn’t voice in the sales process but showed up in the decision.
- Proof points that landed: the specific evidence that moved them.
Feed this into your MSD/MPF. The narrative should evolve as the research does. If you’re updating your ICP and personas but not your messaging, the work is half done.
AI in the Research Workflow
AI is excellent at synthesising large volumes of interview transcripts and surfacing themes. Feed it 10 call transcripts and ask it to identify the three most common problem statements, the language used to describe success, and the objections that appear more than once.
It will get you a solid first synthesis in minutes rather than hours. Zero-Trust applies: read the transcripts yourself and verify the themes before you act on them. AI finds patterns. You decide whether those patterns are real.