Teresa Torres on Continuous Discovery
Teresa Torres in conversation with Lenny Rachitsky. Torres covers the opportunity solution tree, story-based customer interviewing, continuous discovery as an always-on system, assumption testing, and the product trio model.
Key ideas
- Opportunity solution tree: a tree visual — outcome at the root, branching into the opportunity space (unmet needs, pain points, desires), then solutions, then assumption tests. The hard part is staying in problem space: 98% of practitioners write opportunities as solutions.
- Story-based interviewing: ask about the last specific instance, not hypothetical preferences. Stories are grounded, rich with context, and surface needs the customer did not know they had.
- Continuous discovery = continuous feedback loops: discovery and delivery run in parallel at all times, not in phases. Even one interview a week is enough to maintain momentum; the goal is habit, not project.
- Assumption testing over experiments: break an idea into its underlying assumptions, prioritise them, and test each cheaply. A well-functioning team runs half a dozen assumption tests across three concurrent ideas in a single week.
- Product trio: product manager, designer, and engineer decide together from shared understanding. Disagreement signals a missing option, not a hierarchy question to resolve.
Opportunity solution tree
The tree has four levels:
- Outcome — the business or product metric the team is moving.
- Opportunity space — unmet needs, pain points, and desires drawn from customer stories. Structured using an experience map: the chronological steps of a customer’s experience become the top-level branches. Pain points and desires nest below each step.
- Solutions — candidate features or designs that address a specific opportunity.
- Assumption tests — experiments that evaluate whether a solution’s key assumptions hold.
Why it is hard in practice: The opportunity space must be kept in problem language, not solution language. Teams instinctively jump to solutions. Well-structured opportunities are specific enough to solve; poorly framed ones (e.g., “make this easier to use”) are bottomless.
Story-based interviewing
The single most common interview mistake is asking direct, decontextualised questions. Human memory outside a specific context produces stated preferences that do not match actual behaviour.
The fix: anchor every question in a recent specific instance.
“Tell me about the last time you watched something on a streaming entertainment service.”
From that anchor, the interviewer only needs curiosity: “Set the scene. What did you do next? What happened after that?” The technique:
- Situates the interviewee back in the real moment.
- Surfaces needs the customer is not consciously aware of.
- Keeps the interviewer listening, not preparing their next question.
- Makes interviews feel like natural conversation.
A second failure mode is staying shallow. When an interviewee reveals friction, the interviewer must probe rather than move on. That friction is the goldmine.
Continuous discovery
Continuous discovery means running discovery in parallel with delivery at all times. It is not a phase. One interview a week is sufficient on the opportunity-identification side.
On the “no time for discovery” objection: leaders who say this mean “no time for project-based research”. The answer is not to stop delivery and do discovery; it is to run both at all times. Everything in a backlog is a bet, whether or not discovery was done. Discovery makes the bets better over time.
Automating customer access
To interview every week without manual recruiting overhead:
- Consumer / B2B end-users: embed a lightweight intercept in the product. If yes, direct to scheduling software.
- Buyers and decision-makers: route through internal teams (sales, account management, support) already on the phone with those people. Define a recruiting trigger each week.
Assumption testing
| Dimension | Full experiment | Assumption test |
|---|---|---|
| Scope | Tests the whole idea | Tests one underlying assumption |
| Duration | Weeks | Days or less |
| Cadence | One at a time | Half a dozen to a dozen per week |
Work with three candidate solutions simultaneously. Break each into its assumptions, prioritise by risk, and test the riskiest assumption first. By the end of a week, teams can compare three solutions on evidence — not opinion.
Product trio and collaboration
The canonical continuous-discovery team unit is the product trio: product manager, designer, and engineer. They share discovery work, share context, and make decisions together from shared understanding.
When a well-functioning trio disagrees, it means no good option has been found yet — they keep looking. Disagreements reduce dramatically when the trio operates from shared understanding.