☝🏻️ Hard to read? Find the same content below 👇🏻
Another data science concept that I find applies to product management is “overfitting”.
A common trap in both data science and product management but data scientists hear about it all the time and are constantly reminded to avoid the pitfall.
“Overfit” is effectively paying too much attention to your data.
If you try to account for every data point, you’ll end up with a model (which represents your interpretation of the data into a reusable formula or relationship) that works really well for the data that you have but terribly when you bring in new data.
In data science, you can set aside some of your data and exclude it from informing your model. Once you’ve determined your model you can check the fresh data against your model to see how well it performs.
In product, it’s not that simple because we largely live in the world of qualitative data.
Overfit in product pretty much boils down to the cliche of “if you try to be everything to everyone, you’ll end up being nothing to no one”.
Awkward grammar aside, I put a lot of weight in this adage. Especially when you’re in the discovery stage, beware overfitting.
All input is valuable, but it’s not all applicable.
Beware building something that’s trying to be so many things, it ends up being mediocre at all of them.