In one sentence
You cannot reliably predict the future, but you can improve your odds by combining a clear initial strategy with rapid feedback, willingness to revise, and structural defenses against uncertainty. The goal is not to identify one correct plan in advance; it is to create a system that can discover and exploit success as reality unfolds.
Overview
Cohen begins by showing how badly experts forecast markets, drug sales, corporate growth, chess tournaments, and sporting outcomes. Even base rates can fail: Tesla dramatically exceeded historical growth patterns. Product success is similarly emergent: Game Neverending led to Flickr, Glitch led to Slack, and WhatsApp evolved from status updates into free messaging. These stories support a middle path between rigid planning and randomness: start with a thoughtful direction, launch, observe actual behavior, and upgrade the strategy.
Core ideas
Prediction is a weak foundation for strategy
Expertise, sophisticated models, incentives, and historical averages do not make complex systems reliably predictable. Forecasts should therefore be treated as hypotheses or scenario inputs, not as facts on which a plan depends.
A strategy is necessary even when it is wrong
A strategy provides direction, coherence, and enough focus to build something distinctive. Early users and experiments then reveal what the strategy missed. Moving and learning can be more valuable than prolonged deliberation.
Customer behavior upgrades the strategy
Customers are not always literally right about requested features, but their behavior is directionally informative. Especially valuable are unexpected uses or “abuse” of a product: they reveal a stronger underlying need than the original positioning did.
Build plans around optionality
Resilient plans use “or” rather than “and.” Multiple customer segments, channels, pricing models, products, suppliers, or technical approaches mean that one failure does not become fatal. Optionality reduces the need to predict correctly.
Use strengths and structural advantages
Choose battles where your capabilities, knowledge, network, relationships, and accumulated assets give you an advantage. Build moats that remain useful despite changing competitors and markets.
Be intentionally reactive
Replace long planning cycles with frequent delivery, feedback, hypothesis updates, and reversible decisions. This is not aimlessness: a durable long-term purpose guides rapid responses to immediate evidence.
Manage uncertainty through portfolio logic
Hedged bets, redundancy, multiple vendors, parallel solutions, multiple brands, and self-disruption trade some efficiency or maximum upside for more reliable outcomes. The point is to avoid dependence on a single forecast.
Choose better uncertainties
Extreme novelty can eliminate direct competition and create large upside, though the risk must justify the potential reward. Coalitions, standards, and partnerships can also make an organization more resistant to external shocks.
Practical takeaways
- Write down the strategy you currently believe, but label its assumptions explicitly.
- Launch the smallest meaningful version quickly enough to expose real customer behavior.
- Look for customers using the product in an unintended way; investigate the underlying job rather than merely implementing the requested feature.
- For every critical assumption, ask: what alternative route could still produce success?
- Prefer markets, products, and channels where your existing assets create an unfair advantage.
- Use pre-mortems and a futures-cone exercise to identify plausible disruptions and signals that a supposedly unlikely scenario is occurring.
- Reduce unnecessary dependencies and complexity; simple products and operating plans are easier to adapt.
- Invest heavily in durable needs—such as lower prices, speed, reliability, security, or free communication—rather than trying to forecast every trend.
Caveats and counterpoints
- The article’s examples are mostly entrepreneurial success stories, so they may underrepresent the many experiments that pivot repeatedly without finding a viable business.
- Cohen’s argument against base rates is rhetorically powerful but can be overextended. Tesla’s exceptional outcome does not invalidate base rates; it illustrates that outliers occur and that forecasts should combine reference classes with case-specific evidence.
- Optionality, redundancy, and hedging are not free. They consume money, attention, and organizational capacity, and excessive parallelism can create indecision or dilute focus.
- Being reactive works best when feedback is fast and observable. In regulated, capital-intensive, safety-critical, or long-cycle environments, waiting for customer behavior may be insufficient.
- “Customer behavior is directionally right” still requires interpretation. Users can request local optimizations, follow fashion, or resist valuable innovations they cannot yet imagine.
Questions worth revisiting
- Which part of my current strategy depends on the largest number of things all going right?
- What are customers doing with the product that I did not intend?
- What enduring customer desire could remain important even if technology and competitors change?
- Where do I have an existing advantage that should determine which market I enter?
- What low-cost experiments would most quickly falsify my current assumptions?
- Which risks deserve redundancy or insurance, and where would that cost more than the risk is worth?
Return to this when…
Return to this note when making a strategy, deciding whether to keep planning or launch, interpreting unexpected customer behavior, or evaluating whether a plan is robust to several plausible futures. The central reminder: do not confuse uncertainty with helplessness—design the organization so that it can learn, adapt, and still win when its initial prediction is wrong.