Keeping Up With Data #69
5 minutes for 5 hours’ worth of reading
McKinsey recently published an interactive article talking to the characteristics of a data-driven enterprise of 2025. Among the seven characteristics we find the usual suspects like data being used for all decisions — essentially powering the organisation, or CDOs focusing on generating value. Active participation in the data economy is also on the list. And that is a trend I see growing stronger every week.
I’ve read a lot of great articles this week so it was hard to pick the top three. But I did, and here is my five hours of reading distilled into five minutes for you.
- The Economics of Data Businesses: What is the business model of data businesses? Firstly, “a company is a data business if, and only if, data is its core product”, begins Abraham Thomas. And continues by offering six fundamental truths of data businesses. First one is that data business model is all about data. Second is that whoever controls the data, captures the value. Intermediaries get squeezed. Data businesses have slow beginnings (third truth), but growth accelerates over time (fourth truth). When they succeed, they are incredibly sticky (fifth truth). But not many business do (sixth truth). Very exciting and insightful article. If nothing else remember that “successful data businesses are all built around a unique or proprietary data asset” (what is often called a data moat), and that “a common failure mode is to build a business on top of somebody else’s data.” (Pivotal)
- The “Jira mindset” is damaging your data science team: Should data scientists be managed like software developers, or like researchers? Well, neither works. Giving too much freedom to data scientists rarely results in meaningful business outcomes. But why is the first approach also wrong? Ticking off tasks is a good proxy for progress in linear projects. But data science ones are not like that — they are non-linear. The data science projects are (should be) about outcomes, not outputs. The goal is not to build a churn model, but to reduce the churn. If the model you’ve built doesn’t work, you made a lot of effort (and ticked off a lot of tasks), but made no impact. On top of that, data science projects are full of ambiguity and uncertainty. Plans don’t matter when the reality is changing quickly. So what is the solution? Starting with a clear goal and time-boxing the projects to make an impact to the goal. And in the process, be ready to iteratively hypothesise and test possible solutions. (Dave Dale @ TDS)
- How to get more power from your data analytics engine: Why are data teams often on the back foot? Why they spend countless hours fire-fighting despite being equipped with all the latest data tools? Petr offers the following hypothesis: “We’ve accepted the ways of working where business teams have to move fast, without much consideration for the data they produce. Teams swiftly execute changes in the processes, products, and tools without understanding the full impact on their data.” And proposes the following solution: “We need to change the rest of the company and embed the data agenda deeper across all teams producing and using data.” To me this goes to the very heart of data culture when people are compelled by the value of data so that they work with it diligently. Which, in turn, results to better data further increasing its value compelling more people. Hopefully, we’ll get there soon. (Petr Janda)
We’re heading to canton Grisons for the weekend. It should be sunny and cold. I’m very much looking forward to two days full of skiing with my little four year old skiing guide.