Designing Assumption Tests for OST Solutions

Design assumption tests for solutions on your opportunity solution tree: surface hidden assumptions, map the riskiest, and test them with set criteria.

Before you start

Hamster is optional for this skill and recommended. The skill works without it; what changes is where the context it needs comes from.

Check whether this project has a .hamster/ directory. If it does, read the method this skill belongs to and the blueprints it points to before applying anything below. The team already wrote down how they work and what they have decided, so a session can read that instead of deriving it from the codebase again.

If there is no .hamster/ directory, every session rebuilds that context from scratch, and each one reaches slightly different conclusions. Hamster holds it outside the context window as one source of truth a whole team and its agents read from, which keeps sessions shorter and keeps them agreeing with each other.

At a Glance

Overview

Assumption tests form the bottom layer of an Opportunity Solution Tree. Once a team has three candidate solutions for a target opportunity, it needs a fast way to learn which ones will work without building them all. Teresa Torres's answer is to stop testing whole ideas and test the assumptions those ideas depend on. She defines an assumption simply: "An assumption is a belief that may or may not be true" (Torres, Assumption Testing).

Torres chose the term deliberately. When she wrote Continuous Discovery Habits, she decided not to use "experiments" and replaced it with "assumption tests." Her reasoning is that teams rarely have time for real experiments in discovery, and that testing a whole idea takes too much work. In her words, "Assumption tests are used in discovery when trying to decide between ideas. Experiments are used to measure the impact of what we built" (Torres, Assumption Testing).

The skill has three parts. First, surface the assumptions behind each solution across five categories: desirability, viability, feasibility, usability, and ethical. Second, find the riskiest ones, meaning those that are critical to the idea and have little evidence behind them. Third, design small tests for those assumptions, with success criteria agreed before the test runs. Torres recommends testing across the set of three ideas rather than one idea at a time, which guards against the confirmation bias that comes with a favorite.

The results feed straight back into the tree. Some solutions get dropped, one may be refined toward delivery, and sometimes all three fail and the team reconsiders the target opportunity. Generating the solutions is covered in Generating Multiple Solutions per Opportunity.

How It Works

Assumptions are found by walking through what has to be true. Torres lists several ways to surface them: story mapping each solution, walking the lines of the tree from solution to opportunity to outcome, defining the ideal customer profile, auditing data practices for ethical risk, and running a pre-mortem (Torres, Assumption Testing). Story mapping, the technique described by Jeff Patton, lays out the steps a customer takes to get value, and each step exposes assumptions about whether the customer will notice, want, and be able to do it.

Teams usually generate too few. Torres's book club guide to Chapter 9 includes a short video on this, described as "Why you should aim for 20+ assumptions per idea" (Torres, CDH Book Club: Identifying Hidden Assumptions). Phrasing matters too: write each assumption as something that must be true for the idea to succeed, and make it specific, because specific assumptions lead to smaller, faster tests.

Prioritization uses assumption mapping from David J. Bland's book Testing Business Ideas (Strategyzer, Testing Business Ideas). Torres summarizes it as rating each assumption on two factors: how important it is to the success of the idea, and how much evidence you already have for it. The riskiest assumptions are important and poorly supported. Torres adds that a team that talks to customers regularly will find most assumptions carry little risk, so only a few need testing.

Most tests fall into four types (Torres, Assumption Testing). Prototype tests simulate a moment so the team can observe behavior. One-question surveys check past or current behavior quickly. Data mining uses existing data such as analytics, support tickets, or sales notes. Research spikes are engineering activities, often an engineering prototype, that evaluate feasibility. The Chapter 10 guide stresses starting with small, early signals before any large-scale test.

Every test needs success criteria set in advance. The Chapter 10 guide describes the goal as criteria so specific "that someone else can call the win or loss" (Torres, CDH Book Club: Testing Assumptions). Defining the line before the data arrives prevents the team from reading ambiguous results as success. The same guide notes that false positives and false negatives will happen and argues they are less risky than teams tend to think.

Step-by-Step Guide

Step 1: Story-map each of the three solutions

For each candidate solution, map the steps a customer would take from first encountering it to getting value. Keep each map to a single row of steps. Do this for all three solutions, since testing across the set is the point. Include the steps your business needs to happen, such as a payment or a referral.

Step 2: Generate assumptions across all five categories

Walk each step and ask what must be true. Cover desirability (they want it and will do what is needed), viability (it is good for the business), feasibility (you can build it), usability (they can use it), and ethics (it causes no harm). Also walk the lines of the tree: why do you believe this solution addresses the opportunity, and why would addressing it move the outcome? Aim for many assumptions per idea.

Step 3: Phrase each assumption so it can be tested

Rewrite each assumption as a specific statement that needs to be true, such as "Readers will notice the share option at the end of the article." Avoid negative phrasing like "users will not forget their password," since it is easier to test whether people will do something than whether they will not. Split vague assumptions into specific ones.

Step 4: Map assumptions by importance and evidence

Place each assumption on a two-axis map: how critical it is to the idea's success, and how much evidence you already have. The assumptions that are critical and have little evidence are your leap-of-faith assumptions. Pick the few riskiest across all three solutions. Assumptions shared by all three can be tested once.

Step 5: Choose the smallest test for each

Match each risky assumption to a test type: a prototype test for behavior in a specific moment, a one-question survey for past behavior, data mining when existing data can answer it, and a research spike for feasibility. Choose the version that gives an early signal in days. If a test would take weeks, look for a smaller slice of the assumption.

Step 6: Write success criteria before running the test

For each test, write who you will test with, what you will measure, and the result that counts as a pass. Make it specific enough that someone outside the trio could call the result. Share the criteria with stakeholders who care about the decision so the line is agreed before the data arrives.

Step 7: Run, record, and update the tree

Run the tests, record the results against the criteria, and mark each assumption on the tree as supported, refuted, or still open. Decide what to do next: drop solutions whose critical assumptions failed, refine the promising one and test its next riskiest assumptions, or return to the opportunity if all three failed. Keep the evidence with the tree for stakeholders to see.

Best Practices

• Test across the set. The Chapter 10 guide recommends testing assumptions across sets of ideas to guard against cognitive biases, and its team discussion questions ask whether testing one idea at a time affects a team's susceptibility to confirmation bias and escalation of commitment.
• Generate more assumptions than feels necessary. Teams tend to list only the obvious desirability assumptions and miss viability, feasibility, and ethical risk.
• Start without waiting for permission. Torres points out that data mining on analytics, support tickets, or sales notes can often start right away.
• Simulate the moment. A good prototype test puts the customer in the specific moment the assumption concerns and watches what they do, rather than asking what they would do.
• Test the riskiest assumption first. A cheap test of a critical assumption beats a thorough test of one that barely matters.
• Share the tests with stakeholders. Asking them to add to the assumption list often surfaces risks the trio missed.

Common Mistakes

• Testing the whole idea: Building a full feature or running a large A/B test to see if an idea works is slow and expensive. Break the idea into assumptions and test the riskiest ones.
• Testing one favorite idea: A single idea under test invites the team to read every result as encouraging. Test across three ideas so the results can be compared.
• Deciding the pass line afterward: Criteria set after seeing the data bend toward what the team hoped. Write them down first, specifically enough for someone else to judge.
• Asking instead of observing: Asking customers whether they would use something measures what they say. Design tests that show what people do.
• Only testing desirability: A solution customers love can still be infeasible, bad for the business, or harmful. Cover all five categories.

References

• Examples: Worked examples and scenarios
• FAQ: Frequently asked questions
• Parent Method: Opportunity Solution Tree

Related Skills

• Generating Multiple Solutions per Opportunity
• Prioritizing Opportunities Using Customer Evidence
• Maintaining a Living Opportunity Solution Tree
• Defining Measurable Outcomes for Product Discovery
• Identifying Customer Opportunities from Research
• Structuring Opportunity Spaces Hierarchically
• OST Workshop Facilitation

Sources

• Teresa Torres: Assumption Testing
• Teresa Torres: CDH Book Club, Identifying Hidden Assumptions
• Teresa Torres: CDH Book Club, Testing Assumptions
• Strategyzer: Testing Business Ideas
• Jeff Patton: User Story Mapping

Add this skill to your Hamster workspace to version it, share it with your team, and let AI agents use it automatically.