Why Growth Needs to be Model-Driven, Not Data-Driven
Businesses that are data-driven will lose to those that are model-driven
Data-driven decision making means you’re always behind the market, and slow to respond
Model-driven decision making, being quick to respond to the market, and rapid iteration of growth concepts will enable you to get ahead of the market
Update July 31 2020. See this brilliant video by (scientist) Bret Weinstein: Data-driven vs hypothesis-driven science
“You have to build the ark before the rains come.”
My clients make big jumps in creating growth (as opposed to incremental growth) because I teach them to be model-driven – not data-driven – in their decision making. I suggest this approach because data-driven decision making has two, big flaws:
Your data are always incomplete
You will either be paralyzed by inaction, or will always be late
Most people get turned off when I argue against data-driven decision making. Either because they don’t understand what I’m talking about, and/or they’ve bought into the propaganda of “big-data”, and the use of data-driven, evidence-based decision making.
Fortunately (but also unfortunately), I have an example that everyone can agree on: the Covid-19 pandemic. In real time we’re seeing how governments and data “experts” have made – and continue to make – disastrous decisions because they are data-driven, instead of model-driven.
This article will discuss:
Decision making: model-driven vs data-driven
How, in real time, we’re seeing just how wrong data-driven and evidence-based decision making can be
Why real-world decision making should be model-driven, not data-driven
How this relates to creating growth
Model-Driven vs Data-Driven
The differences between model-driven and data-driven can be summed up as:
Data-driven (evidence-based) decision making is about proving what does or doesn’t exist. It deals in certainties.
For example, the scientific method is a data-driven approach. You have something you want to investigate (e.g. did humans come from apes, how fast do things fall to the earth,...), you do a study, gather data, and determine what is or isn’t true. End of story.
Model-driven decision making is about conceiving probabilistic outcomes, and then taking action to realize or avoid one or more of those outcomes. This happens by first developing an understanding of the system; then, when faced with making a decision, you consider the state of the system, calculate various outcomes, and then take action.
For example, given what I know about my 3-year-old daughter, if, at 7pm, I tell her it’s time to take a bath, I project she will:
Cry “I don’t want to take a bath” and run away (most likely)
Ask “Will you or mommy take a bath with me?” (not very likely)
Ask “Can I bring a toy?” (not very likely)
Say “I love baths!” and then run into the bathroom (extremely unlikely)
Based on these outcomes, I choose a course of action (e.g. just tell her it’s time for a bath, offer her a cookie if she takes a bath, or skip the bath).
The chart below shows differences between the two approaches:
How data-driven and & evidence-based killed (and will kill) a lot of people
I was inspired to write the article because of today’s statement by President Donald Trump. He faces a decision:
Don’t shut down the economy, see how bad the Covid-19 outbreak is, and then determine what to do.
Shut down the American economy (at tremendous cost), and take preemptive action that prevents an outbreak.
Strategy A is a data-driven decision because he wants evidence (data) that prove Covid-19 is a deadly outbreak within the US, BEFORE he will take action.
Strategy B, on the other hand, is a model-driven approach. It looks at the behavior of the virus, determines possible outcomes, and then takes action to avoid and / or secure one or more outcomes.
President Trump is certainly not alone in using a data-driven approach to decision making. Famed statistician John Ioannidis, on March 17, suggested that reaction to the Covid-19 pandemic was overblown, that it’s likely to be a “once-in-a-century evidence fiasco” and that “we are making decisions without reliable data”.
He suggested, similar to Trump, that instead of taking aggressive measures aimed at slowing down the outbreak, we should be data-driven and wait for more evidence before taking action.
In other words, Trump and Ioannidis both want evidence of mass infections and deaths, before believing that there will be mass infections and deaths.
Before visiting the absurdity of that last sentence, let's go back in time to January 2020, and see what the World Health Organization (WHO) was saying.
Update (April 15): See The Huge Cost of Waiting to Contain the Pandemic
Today, we can see how terribly wrong these statements were. The reason is because, as stated on January 30, the WHO said that decisions should be “evidence-based”.
Well, we got the evidence alright. As of today (March 22), we have 377,300 infections and 16,520 deaths in 195 countries – that we know of. And as I sit here, in my NYC apartment, the number of (known) infections in NYC is doubling every 2.5 days. You can probably imagine what I have to say to the WHO, Trump, and Ioannidis.
The problem with being data-driven and evidence-based in the real world
Being data-driven and evidence-based is great when you’re improving a manufacturing process or testing the efficiency of a new drug. It works for two reasons:
You know of and have control over all the data. Laboratories and factories are static, isolated environments. As a result, you know everything that is inside them. You know this, because you a) built the environment and b) decided what to put inside of it.
Conclusions are always true. Because these are static environments, what you determined to be true will always be true. If your metal press machine has a broken widget, it will always be broken until you fix it. If a new drug cures a disease in a bunch of mice, that will always be true.
But all that goes out the window when dealing with open, complex, adaptive systems… which is where real life (and markets) happen.
To illustrate, let’s take the WHO tweet from January 14 – where the claim is made that there’s no evidence of human-to-human transmission. There are two issues with this:
Their evidence was incomplete. There was human-to-human transmission happening, but their studies didn’t find data to support it. And because they didn’t find those data, they assumed they didn’t exist.
All conclusions have a shelf-life. Yes, the virus started as only animal-to-human transmission, but that doesn’t mean it won’t mutate into a virus that is optimized for human-to-human transmission. Which had already happened.
The WHO was gloriously wrong because they took an evidence-based approach to their Wuhan investigation. Like studying the results from a lab experiment. But it wasn’t. The WHO was studying something happening in the real world: a complex, adaptive system.
Why Data-Driven Decision Making Isn’t Good for Creating Growth
So what does this have to do with growth? Well, let’s compare what’s happening with the US government’s response to Covid-19, with RIM’s response to the iPhone.
The US government’s response to Covid-19 and RIM’s response to the iPhone (the z10), are two examples of how data-driven decision making is wrong when dealing with complex, adaptive systems.
In both cases, the decision makers “waited for the data” to tell them what to do. And in both cases, it resulted in disaster. Why? Well, as stated before, even if your data are correct (remember, insights will always have some degree of error), you will always be behind the system you're studying.
“Customers know what they want... until they want something else.”
Why? Because the system is continuing to change. By the time you’ve gathered data and drawn conclusions, those data you originally collected are out of date. And even worse, after you’re sat around in meetings, debated what to do, came up with a solution, and delivered a solution to market… you’re REALLY behind.
The market isn’t going to wait for you to catch up, and neither will a virus.
Using Model-Driven Decision Making to Create Growth
So what’s the answer? When making decisions about what to do under uncertainty and risk, should we, as John Ioannidis suggests, just wait for the data and be evidence-based?
Well, if you do that, then you’re either paralyzed by inaction, or you’re always late. And being late in business and markets means your competition is winning customers while you’re still figuring out an offering – or even worse – selling what consumers used to want… just like what happened to RIM.
The answer I suggest – and is a key ingredient to our clients’ success – is to be model-driven. In other words you need models that reflect how markets work, how consumers shop, how they make decisions, how demand is generated, etc.
Then, when you want to investigate a growth opportunity, you put data (from the system’s current state) into these models, use those model-outputs to create growth concepts, simulate them in the market, and generate predictions about how effective each might be at attracting consumers (see Growth Concepts Sprints).
I say might because the honest truth is that we’re dealing with complex, adaptive systems. The best you can do is to:
Be good at generating and initiating growth opportunities
Be quick to adapt and iterate on the opportunities that work, and quick to abandon what doesn’t
Get better at calculating probabilistic outcomes, and take actions which you believe will maximize the outcomes you want, and minimize the ones you don’t.
Most business leaders – and politicians like president Trump – want certainty before making a hard decision. But in life, you get certainty only after the fact. And in business, markets, and pandemics, that is too late.
