Have you ever watched a product team agonise over a feature decision? There they stand, drowning in data and opinions, trying to reach an impossible level of certainty before moving forward. Chaos ensues — analysis paralysis sets in, momentum stalls, and what should have been a straightforward decision becomes a quagmire of second-guessing, decision churn and fear.
The Temptation of False Certainty
It’s easy to see why teams fall into this trap. On the surface, waiting for certainty feels safer. “Why move forward when we could just get more data?” But this drive for certainty is a fallacy — the future is inherently uncertain, and treating it otherwise is a recipe for paralysis.
The Problem with Oversized Decisions
When product teams chase certainty, they end up with:
Endless analysis cycles
Missed market opportunities
Mounting backlog pressure
Unmet stakeholder expectations
The dreaded “just one more data point” syndrome
The Art of Probabilistic Thinking
Here’s where things get interesting. Just as successful investors don’t require certainty to make good decisions, smart product teams can learn to embrace uncertainty. Instead of depending on gut instincts or chasing unattainable certainty, they can simply assess the likelihood of different outcomes and make calculated bets.
Expected Value (EV): Your Secret Weapon
Think of every product decision as a bet where you’re weighing possible futures. It’s not about being 100% right — it’s about making decisions where the potential upside, multiplied by its probability, outweighs the downsides.
Let’s make this concrete. Imagine you’re deciding whether to build that new AI feature based on a dictate from your CTO. After research, you estimate:
- 50% chance it drives significant adoption (+£100,000 in revenue)
- 30% chance it has a modest impact (+£30,000)
- 20% chance it flops (no gain, just development costs)
Using EV, you can calculate the likely return: (50% × £100,000) + (30% × £30,000) + (20% × £0) = £59,000
Now you’re thinking in probabilities, not certainties. This data-driven approach helps you:
Prioritise features based on potential return rather than gut feel (or opinion)
Balance risk against reward with real numbers
Build realistic forecasts that account for uncertainty
Stop chasing perfect certainty and start making smart bets
Making Better Bets: A Step-by-Step Guide
Step 1: Define the Outcome and Its Value
Start by clarifying the intended impact of the ‘thing’ you are considering doing and why it’s (potentially) valuable.
- Define the main objective (e.g., increase adoption by 25%, reduce churn by 10%).
- Understand why this outcome is potentially valuable (e.g., it increases revenue, delivers some strategic benefit, etc.).
- Identify other possible outcomes that could arise from this work, including both valuable and non-valuable scenarios.
- Estimate the financial impact of both valuable outcomes and potential costs or risks (e.g., added costs, lost revenue, increased customer satisfaction). This provides a full view of the potential impact range without confusing costs with value.
Step 2: Narrow Your Ranges
Next, pressure-testing your ranges:
- What would have to be true for the best case?
- What would prevent the worst case?
- What’s the most likely band within our range?
For example:
- Our best features had major marketing pushes; we don’t have that here
- Our floor is 5% because even accidental clicks hit that number
- Most similar features land between 18–22%
Step 3: Assign Probabilities
Turn your ranges into specific scenarios:
- 10% chance of hitting 35% adoption (everything goes right)
- 60% chance of hitting 20% adoption (most likely case)
- 30% chance of hitting only 8% adoption (significant headwinds)
Step 4: Put Real Value on Outcomes
Attach actual numbers to each scenario. For a new premium feature:
Valuable outcomes:
- 35% adoption = £200k/quarter (10% chance)
- 20% adoption = £100k/quarter (60% chance)
- 8% adoption = £30k/quarter (30% chance)
Potential Costs (e.g., additional workflow step that may increase support costs):
- 50% chance of an additional £20k/month support cost if users struggle with the new step.
Step 5: Calculate the EV
Do the math on value:
(£200k × 0.10) + (£100k × 0.60) + (£30k × 0.30) = £89k per quarter
Do the math on cost:
- Expected Support Cost = £20k/month × 0.5 = £10k/month or £30k/quarter
EV: £89k — £30k = £59k/quarter
Step 6: Compare Alternative Bets
Don’t look at one bet in isolation. As Teresa Torres emphasises, we need to make “compare and contrast” decisions rather than “whether or not” decisions. Asking, “Is this feature good?” is the wrong question. Ask, “Which of these options gives us the best EV?”
For each alternative:
- Calculate the EV
- Consider resource and people costs
- Look at timing and market factors
- Think in terms of “better” rather than “good”
Step 7: Size Your Bets Right
Don’t bet the farm on one big feature. Break down your chosen option:
- Split into smaller testable chunks
- Identify your riskiest assumptions
- Keep each bet small enough that you can afford to be wrong
- Plan multiple small bets rather than one big one
Step 8: Make Uncertainty Explicit
Be upfront about what you know and don’t know:
- “We’re 60% confident about our main scenario”
- “Our biggest unknown is adoption rate”
- “We have good data on costs, less on revenue”
- “Here’s where we could be wrong…”
Step 9: Show Your Work
Build trust by being transparent:
- Share your assumptions
- Explain your ranges
- Point to similar past projects
- Acknowledge blind spots
Step 10: Track & Learn
Don’t bet without checking if you’re winning:
- Monitor actual vs. expected outcomes
- Track which types of bets pay off most often
- Identify your blind spots
- Look for patterns in wins and losses
Step 11: Adjust & Iterate
Use what you learn to get better:
- Review outcomes monthly
- Refine your estimation process
- Build team confidence in probabilistic thinking
- Share learnings across teams
The Core Philosophy
Product decisions are bets, not certainties. Success isn’t about being always right — it’s about making smart bets and learning quickly when you’re wrong.
Key Principles
Embrace Uncertainty
— Start with historical data and win rates
— Break big decisions into smaller, testable bets
— Monitor early signals to adjust course
2. Make Explicit Probability Estimates
— Quantify adoption probability
— Define potential revenue ranges
— Assess technical risk factors
— Convert “might work” into actual numbers
3. Size Bets Appropriately
— Keep individual bets small enough to survive being wrong
— Test critical assumptions early
— Maintain a portfolio of bets rather than all-in moves
Building a Learning System
Your rate of learning will become a critical competitive advantage, so invest in it.
Track and Measure
- Document probability estimates
- Record actual outcomes
- Review wins and losses monthly
- Identify patterns in successful bets
Watch for Common Pitfalls
- Overconfidence in estimates
- Ignoring contradictory evidence, confirmation bias and escalation of commitment bias
- Emotional attachment to sunk costs
Taking Action
Ready to put these principles into action? Start by applying these steps to your next product decision. Building a portfolio of smart bets is key to long-term success in an uncertain world. Whether you’re a seasoned product manager or just beginning to explore probabilistic thinking, there’s always room to improve your approach. Commit to smarter, data-informed decisions today — start building your portfolio of small, calculated bets and track your progress toward better outcomes.
If this feels like a significant shift in your thinking, remember: starting small with just one decision is all it takes to begin mastering this approach, so take the plunge:
Choose one upcoming decision
Write down probability estimates
Calculate the EV
Compare and contrast
Make your bet
Track results
Adjust and repeat
The goal isn’t perfect prediction — it’s making incrementally better decisions over time through systematic learning and adjustment. Your journey toward better decisions starts now.