Opportunity Solution Tree

Opportunity Solution Tree

> The connective tissue between a business outcome and the experiments that move it.

What this skill is

A workflow for building an Opportunity Solution Tree (OST) - the discovery framework popularized by Teresa Torres that anchors discovery work to a clear outcome, surfaces opportunities (unmet customer needs, pain points, desires), generates multiple solution candidates, and identifies the assumption tests that need to run before committing. Produces a living artifact a product trio can update weekly.

What it solves

• Discovery work that wanders without a measurable outcome
• Premature commitment to one solution without exploring alternatives
• Building features before validating the underlying opportunity
• Confusion about which assumption to test next
• Roadmaps that conflate outputs (features shipped) with outcomes (customer behavior changed)

When to invoke

• Starting a new discovery cycle for a product area
• Restructuring a quarterly plan that's grown feature-heavy
• Deciding what to research next when multiple opportunities compete
• Onboarding a new product trio to a shared discovery model
• Reviewing why a shipped feature didn't move the outcome

Phase 1: Define the outcome at the top

The root of every tree is a single, measurable outcome - a customer behavior the business cares about.

Outcome statement format: > Increase [metric] from [baseline] to [target] by [date].

Rules for a good outcome:

• It is a behavior, not a feature ("more activations" not "ship onboarding v2")
• It is measurable today (you can see the baseline)
• It has a target - vague aspirations hide trade-offs
• It is owned by the product trio for the cycle

If the outcome is a business Key Performance Indicator (KPI) such as revenue or retention, restate it as the leading customer behavior that drives it.

Phase 2: Map opportunities

Opportunities are customer needs, pain points, and desires - never solutions. Source them from:

• Customer interviews (continuous discovery weekly cadence)
• Support tickets, Net Promoter Score (NPS) verbatims
• Sales call recordings
• Behavioral analytics (funnel drop-offs)
• Internal observation (customer success escalations, churn calls)

For each opportunity, write:

• A short title in the customer's voice
• Source citations (which interview, which ticket)
• The job stage it sits in (using the job map from the Jobs-to-be-Done (JTBD) skill)

Structure the opportunities into a tree:

• Parent opportunities are broad (e.g., "It's hard to know if I'm making progress")
• Child opportunities are specific instances (e.g., "I lose track of where I left off when I switch devices")

A tree with only parents is too abstract to act on. A tree with only children misses the synthesis.

Phase 3: Prioritize opportunities

Score each opportunity on:

• Reach - how many target customers experience this?
• Severity - how painful when it happens?
• Strategic alignment - does solving it support the outcome at the top?
• Confidence - how sure are we, based on evidence?

Use a 2-by-2 matrix of opportunity size × strategic alignment. Pick one opportunity per cycle for the trio to focus on. Multiple parallel opportunities dilute discovery.

Phase 4: Generate solution candidates

For the chosen opportunity, brainstorm at least 3-5 solution candidates. The first solution is almost never the best.

Solution-generation prompts:

• Magic wand: if there were no constraints, what would solve this?
• Borrow from another industry: how does X industry solve a similar pain?
• Inverse: what's the smallest change that could help?
• Service-not-software: what if a human did it?
• Existing-flow: what change to a current flow would do it?

For each candidate, write a 1-paragraph description and a sketch (Loom video, Figma frame, or rough wireframe).

Phase 5: Identify assumption tests

For each solution candidate, list the assumptions that must be true for it to work. Use the four risk categories described by Marty Cagan:

| Risk | Question | |------|----------| | Value | Will customers use it? | | Usability | Can they figure out how to use it? | | Feasibility | Can we build it with available technology and time? | | Business viability | Does it work for legal, sales, support, and finance? |

For each assumption, design the cheapest possible test:

• Survey, prototype test, fake-door experiment, concierge Minimum Viable Product (MVP), A/B test, technical spike
• Define the success and kill criteria before running the test

Run the test. Update the tree weekly.

Phase 6: Tree maintenance

The OST is a living artifact, not a one-time deliverable:

• Weekly trio meeting reviews the tree
• New customer signals add or refine opportunities
• Failed assumption tests prune solution branches
• Successful tests advance solutions toward delivery
• Outcome at the top is revisited each cycle

Output

• One outcome at the root of the tree
• 5-15 opportunities organized parent → child
• One prioritized opportunity for this cycle with reasoning
• 3-5 solution candidates per chosen opportunity
• Assumption-test plan for each candidate with success criteria
• Living artifact (Mural, Miro, FigJam) with edit history

Operating rules

Always

• Anchor the tree on a measurable customer outcome
• Write opportunities in the customer's voice
• Generate at least 3 solution candidates before committing
• Define kill criteria before running an assumption test
• Update the tree weekly with the trio

Never

• Put solutions in the opportunity layer
• Pursue more than one opportunity per trio per cycle
• Ship without testing the highest-risk assumption
• Treat the tree as a static deliverable
• Confuse a feature output with a behavior outcome

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: varunk130
• Source: varunk130/claude-code-skills
• License: MIT

Install and usage instructions live in the source repository linked above.