Using Falsifiability in Innovation Strategies

In science, a theory is treated as false until proven otherwise. It's called a null hypothesis. But in business, beliefs and ideas are often treated as true and important the moment they sound compelling. And that’s how multi-million dollar mistakes happen. Tropicana lost $30 million in two months because executives assumed customers would love their new packaging, but instead they couldn't find it on the shelf. Coca-Cola spent years developing “New Coke,” testing a new flavor instead of asking customers if they wanted something new. Countless companies cherry-pick supporting data while ignoring the warning signs that could save them. Don’t just look for evidence that confirms their beliefs. Use a null hypothesis. Look for ways to disprove your own ideas before the market does it for you. The lesson is simple: Skepticism is not the enemy of innovation. It is the safeguard that prevents ambition from becoming arrogance. If your idea can’t survive a real test, it won’t survive the market. Test it before reality does.

When we trade learning and facts for opinions and likes, we get louder, not wiser. I’ve watched smart teams drift from evidence to slogans without noticing. It starts small, a catchy line beats a careful analysis in a meeting. A viral opinion outruns a data. A neatly packaged “hot take” feels more decisive than a boring chart. It looks efficient. It feels confident. Then reality sends the invoice. Warning signs of that drift: Expertise is dismissed as academic while charisma is treated as a credential. Headlines outrank baselines, and anecdotes beat samples. Metrics are cherry-picked to fit the narrative, not to test it. “We already know” replaces experiments and pilots. Bad news gets reframed as optics rather than information. People who ask for sources are called difficult instead of dependable. Build this instead: Make every major decision a falsifiable bet by asking what would prove us wrong, by when, and where’s the data? Require pre-reads with sources, assumptions, confidence level, and base rates. Run small experiments and let the results move you. Record decisions with the options considered, risks, owner, and the trigger to revisit. Red-team big calls amd assign someone to argue the most likely way we’re wrong. Hold blameless postmortems that end with dated, owned actions tied to leading indicators. Leader behaviors that anchor a culture to reality: Say “I changed my mind because the evidence changed” out loud and often. Reward people who surface inconvenient facts early. Separate message from messenger and ensure the data is evaluated on its merits, not its popularity. Set the norm that skepticism is a service, not sabotage. Publish reversals and what you learned, where you model that truth beats pride. Simple questions that keep us honest: “What would change your mind?” “What’s the verified evidence?” “What’s the real-world case outside our walls?” “What did the experiment say?” If virality outruns validity, the bill always comes due as outages, recalls, rework, attrition, and lost trust. If evidence outruns ego, the payoff shows up as safer bets, faster learning, and fewer surprises. Strong teams don’t fear facts that challenge their opinions. They use them. That’s how you build something that lasts.

Great products rarely begin with a eureka-style idea; they start with half-formed complaints, vague anxieties, and muddled requests that surface in support tickets, sales calls, or corridor chats, and the craft of the PM is to metabolise that raw noise into a single, sharp sentence that every teammate can rally around - one that names a target user, a concrete struggle, and the measurable consequence of leaving it unsolved. The journey from ambiguity to crispness begins by resisting the reflex to translate symptoms straight into features: instead, interview until you can paraphrase the pain in first-person language that earns a customer’s “exactly!”. Triangulate it with behavioural data to confirm prevalence and urgency, then excavate the root cause through a chain of “why's". Next, reframe the insight in a falsifiable problem statement template - “[User persona] cannot [job] because [obstacle], causing [impact]” - and stress-test it with the team: 1️⃣ Can engineering imagine multiple solution paths? 2️⃣ Can design sketch a narrative? 3️⃣ Can analytics propose a success metric? 4️⃣ Could any stakeholder destroy it with a single clarifying question? If the statement survives, anchor it in a living decision document that records supporting evidence, counter-arguments, and the metric that will later prove it solved; if it doesn’t, iterate or discard quickly, knowing a weak problem statement is costlier than no statement at all. Finally, keep the articulation visible - on sprint boards, meeting agendas - because crispness decays over time. New feedback re-distorts the message, new hires reinterpret the goal, and only continual storytelling preserves focus. Mastering this alchemy of distillation not only steers teams away from build-trap solutions but also earns organisational trust because executives fund clarity, engineers code clarity, designers illustrate clarity, and customers buy clarity - everything else is noise!

Black Swans and Falsifiability We like to believe we can predict the future. Forecasts, dashboards, and strategy decks give us comfort. But reality doesn’t always play by our rules. That’s where the idea of Black Swans comes in. A Black Swan is an event no one saw coming, but when it arrives, it changes everything. Think of the 2008 financial crisis or COVID-19. Businesses that thought they were safe found themselves exposed overnight. I came across this idea again while reading The Almanac of Naval Ravikant by Eric Jorgenson. Alongside Black Swans, another concept that stood out to me is falsifiability. Falsifiability simply means that for an idea to be useful, it must be testable. There should be a way to prove it wrong. This is important because strategies always sound convincing when they are presented, but the real test is in whether they can stand up to reality. For example, saying "Our industry will keep growing” sounds good, but doesn’t give you anything you can actually test. A more useful version would be: “If customer demand drops by 30%, here’s how we’ll still manage to survive the next six months.” That’s something you can stress-test, model, and prepare for. Here’s where the two ideas come together. Black Swans remind us that the biggest risks are often the ones we don’t see coming. Falsifiability gives us a way to challenge our own assumptions before the market does it for us. Together, they encourage a mindset of resilience: don’t just predict the future, design your strategies so they can fail safely and still leave you room to adapt. So next time someone tells you, “This is our strategy” or “This is our model,” ask yourself: Is this falsifiable? Can we actually test it to know whether it will work or not? If yes, that’s a strategy worth experimenting with. If no, you already know the answer. #blackswan #Falsifiability

Your theory about why customers churn sounds convincing. You can point to three support tickets that back it up. Your slides are ready. But can you point to evidence that would destroy your theory completely? If not, you don't have a theory; you have a story (and y'all know I love telling a good story)! Karl Popper argued that the mark of a scientific theory isn't that you can prove it right, but that you could prove it wrong. Theories that can explain any outcome aren't theories at all. They're too flexible to be useful. When you debate this theory, you're now playing tennis without a net. This matters at work because we're trained to build cases, not test them. We collect confirming evidence. We run experiments designed to succeed. We avoid the tests that might embarrass us. The result: strategies that sound good but crack under pressure. The fix: make your theories risky. "Shortening our sales cycle to 30 days will close 20% more deals" is falsifiable. "We need to be more customer-centric" is not. Before testing anything, write down what result would force you to kill the idea completely. Then run that test first. Science doesn't advance by proving theories right. It advances by trying to prove them wrong and failing. Your strategy should work the same way. This Week In MIDS: I'm sharing practical insights from my Master of Information and Data Science coursework at UC Berkeley School of Information. More to come.