Principles of Product Management: Discovery Frameworks (Double Diamond, Design Thinking, Continuous Discovery Habits – Teresa Torres)

Discovery Frameworks (Double Diamond, Design Thinking, Continuous Discovery Habits – Teresa Torres)

Discovery Frameworks: A Practical Study Guide

What This Is Discovery frameworks help PMs systematically explore problems, validate assumptions, and uncover user needs before building solutions. They prevent costly "build-it-and-they-will-come" failures by shifting focus from outputs (features) to outcomes (user impact). Example: When Monzo redesigned its onboarding flow, they used Continuous Discovery Habits (Teresa Torres) to interview users weekly, uncovering that new users struggled with "jargon" (e.g., "sort code")—leading to a simplified, visual onboarding that reduced drop-offs by 23%.

Key Terms & Frameworks

• Double Diamond (Design Council): A 4-phase process: Discover (divergent problem exploration)-Define (convergent problem framing)-Develop (divergent solution ideation)-Deliver (convergent solution validation). Think of it as "zoom out, then zoom in" twice.

• Design Thinking (IDEO/Stanford d.school): 5 steps: Empathize (user research)-Define (problem statement)-Ideate (brainstorming)-Prototype (low-fidelity mockups)-Test (user feedback). Key insight: It’s iterative, not linear.

• Continuous Discovery Habits (Teresa Torres): A weekly cadence of interviewing users, mapping opportunities, and testing assumptions to avoid "one-and-done" research. Core tools:

• Opportunity Solution Tree (OST): Visual map linking desired outcomes-user opportunities-potential solutions-experiments.

• Assumption Testing: Prioritize assumptions using ICE Score (Impact × Confidence × Ease).

• Problem Space vs. Solution Space: Problem space = User needs/pains (e.g., "Users abandon checkout because shipping costs are unclear"). Solution space = Features (e.g., "Show shipping costs upfront"). Mistake: Jumping to solutions before defining the problem.

• Jobs to Be Done (JTBD): Framework to uncover the "job" users "hire" a product to do. Formula: When [situation], I want to [motivation], so I can [outcome]. Example: "When I’m commuting, I want to listen to news, so I can stay informed without reading."

• ICE Score (for prioritization): Impact (1–10, user/business value) × Confidence (1–10, % sure of impact) × Ease (1–10, effort to test). Use case: Prioritize which assumptions to test first in discovery.

• North Star Metric (NSM): The single metric that best captures the core value your product delivers (e.g., Airbnb’s "nights booked", Slack’s "messages sent in teams >5 people"). Why it matters: Aligns discovery efforts to outcomes, not outputs.

• Leading vs. Lagging Indicators:

• Lagging: Historical results (e.g., revenue, retention). Problem: Too late to act.

• Leading: Predictive signals (e.g., "users who complete onboarding in <2 mins have 2x higher retention"). Discovery goal: Find leading indicators to test.

• Assumption Mapping: Plot assumptions on a 2x2 grid:

• X-axis: Importance (high/low).

• Y-axis: Certainty (known/unknown). Action: Test high-importance/low-certainty assumptions first.

• Wizard of Oz Testing: Fake a feature’s functionality to test demand (e.g., Zappos manually bought shoes from stores to test online sales before building inventory systems).

• Fake Door Test: Add a button/link for a feature that doesn’t exist yet, then measure clicks to gauge interest (e.g., Spotify tested "Podcasts" in the nav bar before building the feature).

Step-by-Step Process Flow

Scenario: Your team wants to improve retention for a meal-kit delivery app. Here’s how to apply discovery frameworks:

1. Define the Outcome (North Star Alignment)
2. Action: Align on the North Star Metric (e.g., "weekly active subscribers") and a leading indicator (e.g., "users who cook 3+ meals/week have 40% higher retention").

3. Tool: Use OKRs (Objective: Improve retention; Key Result: Increase % of users cooking 3+ meals/week from 20% to 35%).

4. Explore the Problem Space (Double Diamond: Discover/Define)

5. Action: Conduct user interviews (5–10 users who churned) and behavioral data analysis (e.g., "60% of churn happens after the 3rd box").
6. Tool: JTBD to uncover jobs (e.g., "When I’m too busy to plan meals, I want a quick way to get ingredients, so I can cook without stress").

7. Output: Problem statement (e.g., "Users churn because meal planning feels overwhelming after the initial novelty wears off").

8. Map Opportunities (Continuous Discovery Habits)

9. Action: Build an Opportunity Solution Tree (OST):
   • Outcome: Increase % of users cooking 3+ meals/week.
   • Opportunities: "Users forget to order," "Users don’t know what to cook," "Users find prep time too long."
   • Solutions: "Auto-renewal reminders," "AI meal planner," "Pre-chopped ingredients."

10. Tool: ICE Score to prioritize solutions (e.g., "AI meal planner" scores 7×8×6 = 336).

11. Test Assumptions (Design Thinking: Prototype/Test)

12. Action: Run low-cost experiments to validate assumptions:
    • Assumption: "Users will use an AI meal planner if it saves them 10 mins/week."
    • Experiment: Wizard of Oz test—manually send 50 users a "personalized meal plan" via email and track open rates/clicks.
    • Metric: % of users who click "Order this plan" (target: >15%).

13. Tool: Fake Door Test—add an "AI Planner" button in the app and measure clicks before building it.

14. Synthesize and Decide (Double Diamond: Develop/Deliver)

15. Action: Combine qualitative (interviews) and quantitative (experiment) data to:
    • Kill solutions with low signal (e.g., "Pre-chopped ingredients" had low demand).
    • Double down on high-signal solutions (e.g., "AI meal planner" had 22% click-through).

16. Tool: RICE Score to prioritize the MVP (e.g., "Basic AI planner" vs. "Advanced AI with grocery integration").

17. Iterate (Continuous Discovery)

18. Action: Schedule weekly user interviews to refine the solution. Example:
    • Insight: Users loved the planner but wanted "more variety."
    • Pivot: Add a "Surprise Me" option to the planner.

Common Mistakes

1. Mistake: Skipping problem definition and jumping to solutions (e.g., "Let’s add a chatbot!").

2. Correction: Spend 50% of discovery time in the problem space (interviews, JTBD, data analysis). Why: Solutions without problem validation waste resources.

3. Mistake: Treating discovery as a one-time phase (e.g., "We did user research 6 months ago").

4. Correction: Adopt Continuous Discovery Habits—interview users weekly and update the OST. Why: User needs evolve; stale insights lead to irrelevant features.

5. Mistake: Testing solutions with biased methods (e.g., asking users, "Would you use this?").

6. Correction: Use behavioral tests (e.g., Fake Door, Wizard of Oz) instead of surveys. Why: Users often say "yes" to avoid disappointing you but won’t actually use the feature.

7. Mistake: Confusing outputs (features) with outcomes (user impact).

8. Correction: Tie every experiment to a leading indicator (e.g., "If users click the AI planner, retention should increase"). Why: Features-success; outcomes do.

9. Mistake: Ignoring "unknown unknowns" (e.g., assuming you know all user pain points).

10. Correction: Use assumption mapping to identify blind spots. Why: The biggest risks are the ones you don’t know exist.

PM Interview / Practical Insights

1. Tricky Distinction: Discovery vs. Delivery
2. Interviewer Trap: "How do you balance discovery and delivery?"

3. Answer: Discovery is about exploring problems (e.g., interviews, OSTs); delivery is about building solutions (e.g., sprints, roadmaps). Key: Run them in parallel (e.g., dual-track agile—discovery and delivery teams work separately but sync weekly).

4. Leading vs. Lagging Indicators in Discovery

5. Stakeholder Question: "Why focus on leading indicators if revenue is lagging?"

6. Answer: Leading indicators (e.g., "users who complete onboarding in <2 mins") predict lagging outcomes (e.g., retention). Example: Duolingo found that users who completed their first lesson within 1 hour had 3x higher retention.

7. MVP vs. MMP (Minimum Marketable Product)

8. Interviewer Trap: "What’s the difference between an MVP and an MMP?"
9. Answer:
   • MVP: Smallest experiment to test a hypothesis (e.g., Zappos’ fake shoe store).
   • MMP: Smallest product with enough features to deliver value to users (e.g., Zappos’ first real e-commerce site).

10. Why it matters: Discovery tests MVPs; delivery builds MMPs.

11. Handling Stakeholder Pressure to "Just Build It"

12. Scenario: Your CEO wants to add a feature based on a single user complaint.
13. Answer: Use ICE or RICE to prioritize the idea against other opportunities. Script: "Let’s test this with a Fake Door test—if 10% of users click it, we’ll build it in the next sprint."

Quick Check Questions

1. Question: Your team wants to add a "social sharing" feature to increase engagement, but early tests show it hurts NPS. How do you decide?

2. Answer: Kill the feature. NPS is a lagging indicator of long-term retention; engagement without satisfaction is unsustainable. Why: Prioritize outcomes (NPS) over vanity metrics (engagement).

3. Question: A stakeholder says, "We don’t need user interviews—our data shows users drop off at checkout." What’s your response?

4. Answer: "Data tells us what is happening; interviews tell us why." Run 5–10 interviews to uncover root causes (e.g., "Users abandon because shipping costs are hidden"). Why: Quantitative data is incomplete without qualitative context.

5. Question: Your OST shows 3 opportunities: (1) "Users forget to reorder," (2) "Users want more variety," (3) "Users find prep time too long." How do you prioritize?

6. Answer: Use ICE or RICE to score each opportunity. Example: If "forget to reorder" has high impact (users churn) and is easy to test (e.g., email reminders), prioritize it first. Why: Not all opportunities are equal—focus on high-impact, testable ones.

Last-Minute Cram Sheet

1. Double Diamond: Discover-Define-Develop-Deliver. Zoom out, then zoom in twice.
2. Design Thinking: Empathize-Define-Ideate-Prototype-Test. Iterative, not linear.
3. Continuous Discovery Habits: Weekly interviews + OST + assumption testing. Avoid "one-and-done" research.
4. Opportunity Solution Tree (OST): Outcome-Opportunities-Solutions-Experiments. Visualize the path to impact.
5. ICE Score: Impact × Confidence × Ease. Prioritize high-impact, easy-to-test assumptions.
6. Problem Space vs. Solution Space: Define the problem before ideating solutions. Don’t jump to features!
7. JTBD Formula: When [situation], I want to [motivation], so I can [outcome]. Focus on the "why."
8. Fake Door Test: Measure clicks on a non-existent feature to gauge demand. Low-cost validation.
9. Wizard of Oz: Fake functionality to test demand (e.g., Zappos’ manual shoe orders). Build only if users want it.
10. Leading vs. Lagging Indicators: Leading = predictive (e.g., onboarding completion); lagging = historical (e.g., revenue). Discovery focuses on leading indicators!