Most evenings, when I’m not traveling, I make time to get in at least half an hour on my Peloton bike. It helps to wind down at the end of the day. I’ll turn on a show (lately I’ve been slowly savoring this season’s episodes of The Handmaid’s Tale and waiting in anticipation for season 2 of Ted Lasso) and pedal to break a sweat. As my legs push the pedals around, the wheels spin faster and faster until I need to add resistance to slow myself down.
In the physical world, we who have studied physics are familiar with the coefficient of friction—the amount of force it takes for you to push one object against another. When the object is in motion, like the wheels on my bike, that coefficient goes down. Once I’ve gathered momentum, I don’t need to work as hard to go fast.
A flywheel takes this a step further, by accelerating a large rotor to very high speeds and maintaining the energy in the system as rotational kinetic energy. It takes a great amount of effort to start turning the flywheel, but once it’s in motion it builds momentum to keep picking up speed.
In the world of data, the concept of a flywheel is being used to increase customer centricity and satisfaction. My recent podcast guest, Ash Fontana, and the insurance company Lemonade Inc. show us how.
The data flywheel
Any company that has at least dabbled in data analytics or artificial intelligence (AI) knows that it takes time to get started. In order for big data and AI to provide their desired benefits, a company must build a reservoir of customer information. This process can take years and can cause many companies to give up before realizing the effects.
The data flywheel (also known as data network effects) is the idea that as you acquire more customers or users, you’ll obtain more data, which helps you improve your algorithms and ultimately develop a better product to gain more users. The process repeats and the system continues to grow and improve.
The market intelligence company CB Insights explains how it took them four painstaking years to solicit data and populate their reservoir. They turned away from gathering structured data and focused on developing machine learning software to pull from external, unstructured data sources.
Years later, the flywheel began to turn. Customers and companies learned about the power of CB Insights’ data extraction and began to request to submit their data to drive continuous improvement to the system.
The AI flywheel
On last week’s podcast episode, Ash discussed the flywheel concept applied to AI. Ash is one of the most recognized startup investors in the world after launching online investing at AngelList. He’s written the book on becoming an AI-first company.
He explained that, similar to data network effects, the benefits of AI take time to establish. But once the flywheel is in motion, it can only improve.
If you build something that makes a valuable prediction, it will eventually attract customers. These customers generate more data, and your models can learn, retrain, and redeploy. The wheel begins to turn—the system gets better, customers use it more, other customers join, and your accuracy improves. The data loop spins faster, based on the automatic compounding of information.
Lemonade’s flywheel for customer centricity
Ash also referenced the certified B-Corp Lemonade Insurance as an example of how companies are employing AI to delight customers. I’ve been discussing Lemonade as a case study for some time now—they utilize many strategic patterns that align with outthinking the competition, particularly “be good” and community coordination.
Their model follows the same structure as the data flywheel, while throwing customer satisfaction into the mix. The more satisfied customers are, the more they use the system and provide data. Content customers tell their friends about your product, and additional users join. The growth generates predictive data, which improves machine learning, and increasingly accurate predictions further delight customers.
How to identify AI opportunities
You may be saying, “That makes sense for a newer company, but my legacy company hasn’t tried AI,” or maybe you’ve been trying to build a data reservoir and it hasn’t worked. On the podcast, Ash assured that we often think of AI as a “giant brain in the cloud,” but on the simplest level it’s a tool for making decisions.
He recommends the following steps to get started:
- Automation: Consider, what processes do we repeat over and over? How many of these could be automated?
- Prediction: What factors do we want to know about when people might want or need our products? What information do we need in order to predict when customers will want to find us?
Most companies get stuck organizing data or thinking their algorithm needs to be more complicated. Ash warns that there is no “one-size-fits-all” approach to AI and the best way to start gathering data is to begin.