Customer Research as a Documentarian
Observe like a non-judgmental documentarian, find the pattern, then validate the truth with data
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 4
- Confidence
- 90%
A method for doing customer research that actually changes what you build, rather than confirming your biases. Enter conversations as a neutral observer — a documentarian dropped into the jungle — because what customers say and what they want diverge, especially when they're being polite. Separate what they say from what they do, find patterns across many conversations, then validate the truth with prototypes, betas and real numbers, always anchored back to whether it works for the business.
Origin
Rohini Pandhi's articulation of Mercury's 'customer obsessed' research culture; the pricing/value-perception logic draws on Lenny's podcast episode with Monzo's Madhavan (the water pricing example).
Core principles
- 01What a customer says and what they want are two different things — people are especially nice on the phone
- 02Be a researcher, not an advocate: a third-party, non-judgmental observer, not someone seeking validation
- 03One conversation isn't truth — find a pattern heard multiple times, then validate it
- 04Customer obsession must be anchored back to the business; a want that isn't viable (e.g. 'free products') isn't a product
- 05Volume and channels matter: a real culture runs many calls plus surveys, off-boarding signals, and direct social/CS touchpoints
How to run it
- 1
Enter as a neutral observer
Drop your bias before the conversation. Act like a documentarian or third-party researcher watching what's actually happening, not seeking to validate or invalidate a pre-formed view.
Watch out Customers will be polite, blame themselves for UX failures ('I was doing something wrong') and take pressure off you — don't take the flattery as truth.
- 2
Separate what they say from what they do
Focus on understanding what the customer was trying to do and in what context, rather than accepting their stated satisfaction or self-blame at face value.
Pro tip Run high volume when in discovery mode — teams deep in discovery do 20–25 calls, plus surveys for qualitative and quantitative signal.
- 3
Find the pattern, then validate the truth
Look for a truth that recurs across multiple conversations. Then validate it with a prototype, betas, and by checking whether the numbers actually work in your favor.
- 4
Anchor the insight to the business
Triangulate the validated customer need against business viability and pricing context — value perception depends on context (the same water is free at a fountain, costly on a plane). Find where artistry meets the business.
Pro tip Aggregate every touchpoint — CS and RM/sales calls, off-boarding messages in a Slack channel, direct X/email replies — as customer signal, not just formal interviews.
Watch out 'Customer obsession for the sense of customer obsession' fails; a want that can't sustain a viable business (free everything) can't be served.
In the wild
Rohini notes customers on calls will praise a product and say a broken flow was their own fault to relieve your pressure. The documentarian stance treats that as data about their context, not as validation, and cross-checks against patterns and prototype numbers.
→ Research that surfaces real needs instead of flattering, bias-confirming feedback.
Applying the water pricing example (free at a fountain, ~$5 at a venue, more on a plane), Mercury anchors customer wants to business viability and prices complex accounts-payable/receivable features for the ICP who reaches that need 6–18 months into their business.
→ Customer obsession balanced against a viable, willingness-to-pay-tested pricing model.
Common mistakes
Going in with a bias to validate
Seeking confirmation turns research into theater; you hear what flatters you rather than what's true.
Acting on a single conversation
One data point isn't a pattern — treating it as truth without recurrence and validation leads to building the wrong thing.
Customer obsession detached from the business
Chasing every want (customers 'probably want free products') without anchoring to viability builds things that can't sustain the company.
Is it for you?
Best for
PMs and researchers running discovery who want feedback that actually changes the roadmap
Not ideal for
Late-stage optimization work where the need is proven and the task is quantitative tuning, not discovery
From the transcript
“there's a little bit of like what a customer says versus what they want going to be two different things”
“You want to be a researcher. You want to be like this third-party non-judgmental kind of like you're being brought into the jungle and just…”
“get to a truth that seems to like find a pattern or you hear about multiple times and then then validate that truth”
“the customer obsession needs to also be anchored back into does it work for your business?”
From the episode
How to build your product team from scratch, attract top product talent, go multi-product, and more
Rohini Pandhi (Mercury, Square)