How Much Time Should You Spend in Product Discovery?

Teresa Torres argues that asking how much time to spend in discovery is the wrong question, because discovery and delivery should happen continuously rather than in sequence. The right amount of discovery is whatever is enough to mitigate unacceptable risk, and that depends on what is learned along the way. She contends that speed of learning matters more than speed of writing code, and that teams often err in either direction, relying too heavily on building or on research without shipping. To avoid analysis paralysis, she recommends mapping decisions in an opportunity solution tree and classifying choices as one-way or two-way doors, borrowing Jeff Bezos's framing so reversible decisions get made quickly.

01Key takeaways

• Stop treating discovery as a finite phase; do it continuously alongside delivery.
• Judge discovery by how fast you learn, not by how much code you write, and build only when it is the quickest path to insight.
• Map outcomes, opportunities, and solutions in an opportunity solution tree and revisit it whenever new information arrives.
• Classify decisions as one-way or two-way doors, deliberating on the irreversible ones and deciding the reversible ones quickly.
• Be inclusive when mapping opportunities, and use current data to pick a target rather than waiting for perfect data.
• Maintain momentum by shipping regularly, since building the wrong thing and building nothing both fail customers.

02Key sections

Reframing the question

Discovery is not a phase that precedes delivery, and there is no universal answer to how long it should take. The real goal is mitigating unacceptable risk, which depends on what is learned.

Speed of learning over speed of coding

Building code can be a valid learning tool if it is the fastest route to insight, even when the code is discarded. Teams err both by over-relying on A/B tests and by researching for months without shipping.

Discovery as a non-linear process

Outcomes, opportunities, and solutions are explored simultaneously, and new learning should prompt revisiting earlier decisions. Analysis paralysis at each step undermines continuous discovery.

Opportunity solution trees

An opportunity solution tree maps the routes from a desired outcome through opportunities to solutions, letting teams prune, grow, or restart branches as they learn. It differs from a feature roadmap by showing many possible routes to a known outcome.

One-way and two-way doors

Irreversible decisions deserve careful deliberation, while reversible ones should be made quickly by small groups. Most digital product decisions are two-way doors with fast feedback loops.

Bringing it together

Keep balancing discovery and delivery continuously, use the tree and door classification to move fast, and increase discovery effort week over week.

03From the post

“I get asked all the time, “How much time should we spend in discovery?” I have two problems with this question. First, it assumes you do discovery first and then delivery second, which is not true. You should be discovering and delivering all the time. Second, it implies that there”

04Frameworks mentioned

Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.