Agree to Disagree: Data science versus Data Entrepreneurship

Hi, I’m Mark van de Pol responsible for the businesses that work with JADS and for months now I have been having heated — but constructive — discussions with my academic counter weight Arjan Haring 🔮🔨 on Data Entrepreneurship versus Data Science. What’s the difference and why does it matter?
Some time ago we agreed to disagree. But I guess our discussions are not unique, most probably there are tons of people having similar discussions right now. That’s why we thought it would be a good idea to go public with our internal discussion, hoping to move the discussion, the field and the industry forward.

So the first thing I would like to know from Arjan is:

Why do you talk so much about causation? A strong correlation is good enough… Right?

Arjan: That’s a really good question. And I would agree with the statement that the business case for chasing causality is weak. The business case for correlation is much stronger. But I would argue that in a rapidly changing environment you have to be an innovator. And the innovators dilemma, in my opinion, is a trade-off between when should you stop exploring new possibilities and start exploiting the knowledge that you already have.

With correlations (or blackbox solutions) you don’t find out what the confounders are

With correlations (or blackbox solutions) you don’t find out what the confounders are, the real reasons why something happens. So in a market situation, I think that Uber found out one confounder that no one yet understood, people want to be certain their cab is on the way to them. They hate waiting for something that might not even show up.

With these piece of the puzzle (and other pieces of course) Uber was able to disrupt the cab industry. So, I would make the case that the company that understand the markets and its customers better (i.e. causality) has a better chance to survive.

How would you comment on that hypothesis?

Mark: I certainly agree with your hypothesis. In a rapidly changing world, it is important to be adaptive as a company, you have to stay agile. This requires a deep understanding of your customer. You have to be able to tailor your offer and processes to what the market demands. However, there is one important point that should not be overlooked and that is the speed of learning and acting.

Anyone who focuses blindly on in-depth research to arrive at causal relations will be outpaced by reality.

As far as Uber is concerned, it is a great example. They didn’t not come to this connection through scientific research, but rather through continuous real-life experiments. In other words: in an early phase they have checked their offer with potential customers.

Anyone who focuses blindly on in-depth research to arrive at causal relations will be outpaced by reality.

This is how they discovered who their early adopters were, what those customers wanted and adapted their offer accordingly. When they had a clear niche and the business was up and running they could scale up and reach the larger market. Not by saying here you have our product and good luck with it, but precisely by continuing to stay humble in every step of the development, to learn from experiments and to adjust the offer when necessary.

Eventually, key drivers of success became clear.

If every entrepreneur, as far as I am concerned also employees and intrapreneurs, would adopt this mindset a bit more than the world would look a lot nicer.

We would transform our companies from making decision based on gut feeling to decisions that are supported by data. By this I do not mean causal connections, but have a clear view of patterns that can be monetized. For this I do have an effective and manageable ‘experiment hack’ that everyone can apply.

Talk to a group of (internal) customers on a regular basis and ask what should you stop doing, what you should continue doing and what you should do additionally. You will notice how many things you actually overlook and with which you can use to your advantage.

If you are talking to 10 customers during a month and can increase to 10 per week, 10 per day or even each, then you experience how the speed learning will increase. Companies like Uber certainly do 1000 experiments with large groups of customers per day. For that, they also use more advanced techniques.

Let’s go apply it. With your data science perspective. What should business stop, continue and add? I think this will make it even clearer where our thinking differs.

Arjan: great stuff Mark. Wrong, but great. Let’s continue this discussion on my medium account. You don’t really think that I am that naïve that I don’t see what you are trying here…

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.