Hypothesis-First Customer Discovery
Bring a crisp hypothesis to the interview — then be its judge, not its lawyer.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 93%
Duggal argues that most customer research fails in one of three places: no hypothesis, a hypothesis the researcher has fallen in love with, or questions that ask instead of observe. His counter-process is to write a sharp point of view first, then run discovery designed to invalidate it — observing over asking, asking indirectly over directly, and making the customer sell you rather than the reverse. He pairs it with a radical simplification of research logistics: the PM personally phones ten customers rather than commissioning a thousand-person survey.
Origin
Duggal's own synthesis at Nubank, drawing on design thinking, Procter & Gamble's observational research (the Swiffer — people hate mopping but have workaround-adapted to the point they no longer notice the pain), and Jeff Bezos's dictum, cited from his Lex Fridman appearance, that when the data and the anecdote disagree, the anecdote is usually right.
Core principles
- 01Without a hypothesis, research is just an interesting conversation you cannot act on.
- 02Be the judge of your hypothesis, not its lawyer.
- 03The anecdote usually trumps the data — never lose sight of the customer's actual words.
- 04Observe more than you ask; ask indirectly more than directly.
- 05Search for the problem, not for validation of your solution.
- 06Customers can't name innovations they'd love — they've normalised their own pain.
How to run it
- 1
Write a crisp, falsifiable hypothesis before any research
State a specific point of view on what the customer will say and what will make them excited. Without it, you will collect qualitative and quantitative data and have no idea what to make of it, because nothing is being validated or invalidated.
Pro tip A good hypothesis names a pain the customer has stopped noticing because they built workarounds around it.
Watch out Teams routinely skip this step because it feels biasing. Neutrality is not rigour — it's an absence of a claim to test.
- 2
Switch roles from lawyer to judge
Once the hypothesis is crisp, the dominant failure mode is falling in love with it. Explicitly assign yourself the role of judge deciding whether the hypothesis is right, not advocate arguing that it is.
Pro tip Name the disconfirming evidence you would accept before the first interview. If you can't, you're already the lawyer.
Watch out Duggal admits he has fallen into this trap 'too many times to count'. Seniority makes it worse, not better.
- 3
Call ten customers yourself — don't commission a survey of a thousand
Skip the three-week, thousand-respondent survey with cross-tabs that nobody can read at fine grain. Have the PM or designer personally pick up the phone to ten customers. By the first call you can usually predict what customers 6 through 10 will say, and you hear tone of voice — the fine-grain signal statistics never carry.
Pro tip A phone call, not an email — tone of voice carries the emotional intensity that decides whether a pain is real.
Watch out Every layer between the customer and the builder (researcher → plan → summary → readout) destroys signal. Remove the degrees of distance.
- 4
Observe rather than ask; ask indirectly rather than directly
Never ask 'would you love this product?' Approach the same question from multiple indirect angles, and where possible watch behaviour instead of collecting opinions. You are searching for the problem, not the solution.
Pro tip Look for the workaround. Duggal's example: he and his wife opened accounts at different banks purely to organise different life goals — that workaround IS the unmet need.
Watch out Observation is a genuinely tricky skill and hard to do consistently well; Duggal names it the area of greatest inconsistency he still sees in product development.
- 5
Make the customer sell you — play devil's advocate against your own idea
When a customer says something even slightly in the direction you were hoping for, the temptation is to get excited and slip into selling them. Do the opposite: push back against your own idea, argue the other side, and force the customer to convince you the pain is real.
Watch out The moment you start selling, the interview stops producing information and starts producing agreement.
In the wild
Duggal researched the origin of the joint bank account: an artifact of the late 19th / early 20th century, coincident with an era when women could not open an account without their husband's approval. Nubank's hypothesis is that a product 20 to 150 years old is unlikely to fit the modern customer — and that our financial lives are inherently social. Evidence: it is easier to share a Spotify playlist with your wife or daughter than to share a savings goal. The team went into research not to sell the idea but to see whether customers were actually experiencing that pain.
→ A live product development cycle at Nubank around a new social banking arrangement — sharing financial life with spouse, kids, and parents — driven by a strong hypothesis tested adversarially rather than confirmed.
Procter & Gamble's insight was that people hate mopping but will not report it as a problem, because they have developed workarounds and simply accept the burden. Only granular observation of what actually goes wrong reveals the pain — you must bring a strong hypothesis that this IS a real problem even though customers aren't telling you it is.
→ A new product category built on a pain customers had stopped consciously noticing — the canonical case for observing over asking.
Common mistakes
Doing research with no hypothesis in the name of neutrality
PMs often feel they must arrive unbiased. The result is data that neither validates nor invalidates anything, and a conversation that is interesting but not decision-making.
Becoming the lawyer for your hypothesis
After two months of design work, the excited PM asks 'would you love this?' — hunting a yes. That interview produces agreement, not information.
Outsourcing the customer voice through layers of research
Asking a researcher to write a plan, run the study, summarise it and present it inserts so many degrees of distance that the fine-grain signal never reaches the person building the product.
Doing sophisticated analysis while forgetting the question
Duggal repeatedly sees teams run very sophisticated analysis having lost sight of what the question was or what conclusion could truthfully be drawn from the answer.
Is it for you?
Best for
Product managers and designers doing discovery for a new product or a product that has launched but is not yet resonating.
Not ideal for
Large-scale quantitative sizing or pricing studies where statistical representativeness genuinely matters, and mature optimisation work driven by A/B testing.
From the transcript
“teams often skip the step of clearly very very clearly working very hard to be super crisp in articulating their hypothesis”
“I'm a lawyer for my hypothesis I'm not a judge of whether the hypothesis is right or not right”
“observe more than ask questions if you can ask indirect questions more than direct questions”
“make the customer sell you in the interaction and you play The Devil's Advocate”
“don't worry about surveying a th customers taking three weeks surveying a thousand customers getting a bunch of cross tabs”
“you pick up the phone call 10 of them and nine times out of 10 by the time you've made your fist call You can…”
From the episode
Be fundamentally different, not incrementally better
Jag Duggal (Nubank, Facebook, Google, Quantcast)