Insights Don’t Matter
Value comes from action.
The only way to create value, is by doing something. By acting. Information can help you perform better actions, but information alone is useless.
Data functions are currently focused on how to extract, collate, and analyse information, whereas we should be thinking about how data can lead to valuable actions.
Build and maintain infrastructure, then do the things that result in the most valuable actions for the business.
The wrong way to think
People used to think that just collecting terabytes of data would magically transform to increased profitability for their business.
We now know this is wrong - that data has to be organised and accessible. Data was to be a “product” that internal consumers would, well, consume. The whole org would be “data-driven”.
The problem is that this only works if people are using the data intelligently. Dashboards might provide peace of mind, but if they are used in the wrong way, they are useless.
Data isn’t a product. If I’m selling a product, I don’t care what people are using my product for, and how often they are using it, I just care if they buy it (or not). The market will tell me how valuable it is.
But this isn’t true for data - there is no market for determining value. What people say they want, and how they are using data, are very different from the most valuable use of the data.
Some people realised this and decided we shouldn’t be data-driven, we should be insights-driven. Rationale: data in the hands of the data-illiterate is dangerous. People who know how to use the data intelligently should be given control and allowed to investigate anything they think could be insightful.
But insights alone are useless, too. Sure, they may lead to some abstract increase in knowledge about the business, but without actions resulting from them, they aren’t valuable.
Maximise expected value
Here’s an equation:
value = number of actions * value per action
Which explains the following value hierarchy:
Models > Insights > Dashboards
Models (in production and as part of consequential pipelines) can perform actions constantly. Insights usually don’t result in any action at all, but they do at least have the potential to result in high-value change. Dashboards barely get looked at and when they do it’s usually as a sanity check. They rarely result in action.
This is also a helpful way to think about which models to create: those that are part of frequent and/or valuable processes should be prioritised.