Your Best Discovery Calls Are Happening Without You.

This week your company ran forty discovery calls.

You were invited to none of them. They're called sales calls.

While you were recruiting participants for next month's five discovery interviews, the AE ran twelve demos, CS ran nine onboarding sessions, and somebody on the success team had a very honest conversation with an account that's wobbling on renewal. Every one of those calls was recorded. Every one of them was discovery. And in most product orgs, not one of them will ever be read by the person deciding what to build.

An objection in a demo outranks an answer in an interview

Think about what a discovery interview is. You wrote the questions. You recruited the participant. They know they're being studied, and they perform accordingly: thoughtful, agreeable, eager to be useful. You are leading the witness, gently, with a discussion guide.

Now think about what a sales demo is. A prospect with budget on the line watches your product and says "we looked at this last year and passed because the integration story was thin." Nobody asked them that. They brought it up because it's what decides the deal.

That's the difference between prompted and unprompted evidence:

• The interview answer is a response to your framing. It tells you what customers think about what you asked.
• The demo objection is their framing. It tells you what they think about, period.

Unprompted signal carries stakes. People are precise about their pains when money is moving. The renewal call where a customer explains exactly why they're hesitating is the most honest conversation your company will have all quarter, and it doesn't appear in anyone's research plan.

It's not snobbery keeping PMs out. It's format.

PMs don't ignore sales calls because they think the data is beneath them. They ignore them because the data arrives in a shape no PM can use.

The recording is an hour long. There are two hundred of them. The CRM notes were written to track deal stage, not product signal: "good call, sent pricing, follow-up Thursday" tells you nothing about the three feature gaps the prospect named. So the only usable residue of forty discovery-grade conversations a week is whatever fragments survive in a pipeline tool built for forecasting.

Watching the tapes doesn't scale. Trusting the notes loses the signal. So "discovery" quietly shrinks to mean the five calls the PM personally runs, and the other hundred and ninety-five become exhaust.

The genre of the call stops mattering when you read at scale

Here's what changes when every customer conversation, whoever ran it, gets decomposed into structured signals: each objection, pain, feature request, and competitive mention extracted with the verbatim quote, the speaker, and the account attached.

A pain mentioned in three sales demos, two onboarding sessions, and one of your interviews stops being six anecdotes scattered across three teams' tools. It's one pattern with six receipts, counted across unique accounts, traceable to the exact moments it was said. The demo and the interview turn out to be the same research material at different levels of polish.

And the questions you'd never assign a research sprint become askable: which objections are climbing in mid-market demos this quarter? What do churning accounts say in their last three calls that healthy accounts don't? The answer comes back with quotes from calls you never attended, cited down to the moment, including the counter-evidence.

Your sales team has been running your largest research program for years. Nobody was reading the results.

Discovery is a property, not a meeting type

The reframe worth keeping: discovery isn't a genre of call that belongs to the PM. It's a property of every customer conversation your company has, and most of it is happening without you in the room. The teams that get this stop asking "how do we do more discovery" and start asking "how do we stop discarding the discovery we already do."

Five interviews a quarter is a research plan. Two hundred conversations a quarter is a research program. You're already paying for the second one.

---

We built Fragment to read every customer conversation as discovery, whoever ran it, so the objection from Tuesday's demo lands on your desk as evidence instead of dying in a CRM note.