Probabilistic Thinking
A thinking approach that acknowledges uncertainty, assigns probability numbers to each belief, and updates those numbers when new evidence arrives.
Disciplines
Origin Story
Thomas Bayes formulated the theorem to update probabilities based on evidence in the 18th century. Daniel Kahneman and Amos Tversky then showed humans tend to be poor at intuitively assessing probabilities, requiring structured approaches.
Core Principles
- 1Almost nothing is certain; use probability ranges
- 2Express beliefs in numbers (e.g., 70%), reaching for precision beyond vague words
- 3Update probabilities when new information arrives
- 4Use expected value (probability × impact) for decisions
- 5Train calibration: how often 70% predictions are right 70% of time
When to Use
Use in forecasting, risk assessment, investment decisions, business strategy, product development, and any situation filled with uncertainty.
Step-by-Step Guide
Define Question
Write what event or outcome you want to estimate.
Determine Initial Probability
Give initial number (prior) based on historical data or experience.
Gather Evidence
Seek new data, signals, or expert input that's relevant.
Update Probability
Adjust initial number with evidence. Use simple Bayesian logic or adjustment table.
Calculate Expected Value
Multiply probability by impact for each option, choose highest expected value.
Evaluate Calibration
Record actual results and compare with predictions to improve future accuracy.
Probabilistic Thinking
Overview
Probabilistic thinking requires acknowledging that the world holds many shades of grey. Almost everything carries a degree of possibility. By expressing probabilities in numbers, we force ourselves to be honest about uncertainty and can improve decisions as new data arrives.
This approach grounds our judgment in evidence and tempers raw intuition with disciplined estimation. We learn to ask, "How likely is it?" and "How does that likelihood change after the latest information?"
Origin Story
Thomas Bayes developed formulas to update probabilities based on additional evidence. This formula became the foundation of modern analytics from spam detection to product recommendations.
Kahneman and Tversky revealed various biases when humans assess probabilities: overconfidence, ignoring base rates, or trusting dramatic stories more than statistics. Their research showed the importance of more structured approaches.
Core Principles
1. Express Probabilities in Numbers
The phrase "likely" means different things to everyone. 70% is far clearer and comparable.
2. Update with New Evidence
Every new piece of information should change probabilities. If data supports, probability increases. If data weakens, probability decreases. The key is the discipline of updating numbers when evidence shifts.
3. Use Expected Value
The best decision often comes from weighing probability and impact together. For example, 40% chance of getting IDR 10 million (~$645 USD) has expected value of IDR 4 million (~$258 USD), possibly more attractive than 90% chance of getting IDR 3 million (~$194 USD) (expected value IDR 2.7 million/~$174 USD).
4. Calibrate Yourself
Record predictions and results. If you often say "80%" yet land at 40% accuracy, your calibration needs improvement.
Brief Application Steps
- Formulate question: e.g., "Will new feature increase retention?"
- Set prior (initial probability) based on historical data.
- Create simple table: initial probability, new evidence, and adjustment.
- Calculate expected value for each option (e.g., continue project or postpone).
- Record actual results after certain period and compare with predictions.
Case Studies
- Logistics Company: Assesses probability demand increases before holidays by examining three years' data, weather forecasts, and major client promotion schedules. Results: they add temporary fleet only 30% of total to balance probability and risk.
- Weather Analyst: Meteorological agency issues 80% rain forecast for certain area. This number guides local government to prepare additional water pumps.
- Digital Product Team: Before building feature, team makes probability prediction of success, then runs small A/B experiment. Experiment results update probability and determine if feature continues.
Practical Tips
- Use consistent probability scales (e.g., 10%, 30%, 50%, 70%, 90%) for easy comparison.
- Create monthly prediction journal to train calibration. Review quarterly.
- Combine probability assessment with extreme risk discussions. Sometimes small probabilities with large impact still require special mitigation.
Probabilistic thinking makes us calmer when facing uncertainty. Through numbers and evidence, we update decisions gradually, and each new signal sharpens our estimate.
Use Cases
Business Forecasting
Estimate probability of reaching sales targets or campaign success.
→Marketing team gives 60% probability quarterly target is reached after considering sales trends and distribution channel readiness.
Investment
Weigh multiple return and loss scenarios.
→Investor assesses 20% probability a startup succeeds big and 80% fails. Expected value helps determine investment size.
Medical Diagnosis
Update beliefs after receiving test results.
→Doctor recalculates disease probability after lab results come in, drawing on a broader base than the initial symptoms alone.
Product Development
Assess new feature success before allocating large resources.
→Product team gives 30% probability new feature improves retention. Based on expected value, they choose cheap experiment first before building fully.