Roko PlatformDocs

Discovery

Opportunity solution trees

An opportunity solution tree turns interview findings into one outcome, the opportunities under it, and one target. The team builds it before anyone discusses solutions.

An opportunity solution tree connects the result the team wants to the customer needs that could produce it. Teresa Torres introduced it as the working map of continuous discovery. Use it when research exists but the team has not agreed what to build.

An opportunity is something a person said. A solution is something you thought of. The tree keeps the two on different levels, and that separation is the whole discipline.

Outcome

Suppliers finish registration in one sitting

Measured: 40% to 70% of registrations completed in one session

  • Opportunity

    “I don’t have the tax form to hand.”

    7 of 12 interviews

  • Target opportunity

    “I can’t tell what is still missing.”

    9 of 12 interviews

    • Solution

      Checklist of remaining steps

      Test

      Prototype session with 5 suppliers

    • Solution

      Summary page before submit

      Test

      Same session, second flavor

    • Solution

      Reminder email that lists the gaps

      Test

      Send it by hand to 10 stalled suppliers

  • Opportunity

    “I wait days for someone to approve my invite.”

    3 of 12 interviews

Opportunities carry the customer's words and a count of the evidence. The team picks one target, and only then lists solutions and the tests that compare them.

Levels

LevelHoldsExample
OutcomeOne measurable change in customer behaviorSuppliers finish registration in one sitting
OpportunityA need, pain or desire, in the customer's words, with its quote and count"I can't tell what is still missing." (9 of 12)
SolutionOne way to serve the target opportunityA checklist of remaining steps
TestThe smallest thing that compares solutions or checks an assumptionA prototype session with five suppliers

Building the tree

  1. State one outcome

    Write one measurable change in customer behavior, and how you measure it. Do not accept a feature, a launch date, or a revenue figure the team cannot move directly.

    You are givenWhy it failsAsk instead
    "Ship the new editor"It is a solutionWhat would people do differently if it existed?
    "Grow revenue 20%"Nobody can act on it on MondayWhose behavior changes on the way to that number?
    Two outcomesThe tree has two rootsWhich one this quarter?

    The step is complete when the outcome is written with its measure and the team agrees it is the only root.

  2. Lift opportunities from the transcripts

    Read the interview notes. Pull out needs, pains and desires in the customer's words, and attach the quote to each one. "I lose the thread when I come back after a week" is an opportunity. "Add a changelog" is a solution.

    Count the evidence under each opportunity: how many people said it, and whether you heard it or inferred it. An opportunity with one mention is a lead, not a finding.

  3. Structure the tree

    Group and nest opportunities until siblings are real alternatives to each other. Apply two tests at every level:

    • Siblings do not overlap. If solving one partly solves another, the level above is drawn wrong. Regroup instead of removing duplicates.
    • Siblings are complete enough to choose between. If serving every sibling would still not deliver the parent, something is missing.

    Name the places where a test fails. A tidy tree that hides an overlap is worse than a messy tree that shows it.

  4. Pick the target

    Choose one opportunity to work on now, and write down why it beat its siblings. Weigh:

    • Size: how many people, and how often.
    • Urgency: what it costs them today.
    • Fit: whether this team can serve it well.
    • Position: a parent opportunity is safer, and a leaf is easier to act on.

    The choice is an argued judgment, not a score.

  5. Compare solutions for the target

    Only now list solutions, and only for the target. List at least three. A team with one idea compares nothing. Then run the smallest test that tells the options apart, usually a prototype with several flavors, or an assumption test.

Common mistakes

MistakeFix
An opportunity names a featureRestate it as the need underneath
Every opportunity has one quoteYou have one interview's worth of tree. Interview more.
Siblings are the same need at different sizesThe parent is wrong. Redraw the level.
Solutions appear while you lift opportunitiesPark them in a list and return to the transcripts
The tree turns into a backlogA tree holds what people need. A backlog holds what you will build.
The tree never changesUpdate it after every interview. A static tree has stopped learning.
One solution per opportunityGenerate at least three, then compare

Build the tree with your agent

Ask your agent to build a tree from your interview notes. Roko Platform gives the agent the four steps above over MCP, with a completion test for each, and tells the agent to refuse solutions until a target is chosen. The agent needs interview transcripts. If none exist, run customer interviews first.

PromptBuild a tree from interviews
Build an opportunity solution tree from the interview notes in research/. Start by proposing one outcome and wait for me to agree. Quote the source for every opportunity and count the evidence. Report every place where siblings overlap.

The platform does not store the tree as its own object. Keep it where the team works: as a markdown or diagram artifact, or in the change request that acts on the target. When the chosen solution needs evidence before it is built, add a product discovery goal for it on the Roadmap. See Discovery.