Machine Learning in Sales & what your team has in common with Crossfit athletes — Seekers

João Cruz
3 min readMar 16, 2020



What can sales managers learn from the Fittest On Earth? Is there a way that we can look at an athlete’s performance and figure out what his full potential is and what’s hindering his performance? Can we do the same with our sales team? What about the KPI’s we have in place, can we fine tune them to make sure that all of them are pulling in the same direction?


Crossfit is the sport of fitness, in which athletes work their overall competence, and competitive events are devised to single out the overall fittest athlete, the one that excels at it all, from strength and power to speed and endurance. The catch? Usually people who are great in strength tend to lack in endurance, and vice versa. Therefore, when they work too much on one aspect, they’re compromising the other side of the scale.

If you’ve ever used KPI’s to manage a sales team, I’m sure you can relate. Companies start by measuring one KPI, then they add another for a specific aspect of the company’s goals, and one more after that, and so on… Most of the time no one stops to think if any of these KPI’s correlate. So when you get to sales people, they’re often confused about where they should focus their energy. Let’s shed some light on why this happens!


As an example of the kind of analysis we can do with a team’s KPI’s, we’ve gathered data from hundreds of Crossfit athletes and started a Notebook to delve into it.

The first step was to build a Machine Learning algorithm that would learn from the available data and help predict an athlete’s optimal (not maximum, we don’t want to tip the scale too much) results. To accomplish this, a correlation heatmap helps us asses how each KPI relates with another.

The heatmap showed high correlation between Olympic Lifts (Snatch and Clean&Jerk), and Power lifts (Deadlift and Backsquat). The difference is that OL’s are highly technical, hence we can use ML to predict how much an athlete can optimally lift based on his PL’s in order to have the highest overall fitness performance. The algorithm also allows for an estimate of the athlete’s potential from which we can assess if he’s lacking in strength or technique in order to achieve optimal results.

The second step was to build clusters based on strength, endurance, diet and body type, to see if there was any differentiation between types of athletes, and the data shows an incredibly homogeneous distribution, proving that Crossfit athlete’s really are extremely versatile, even across weight classes. This validates our initial premise: we assumed that the overall dataset would be a good training set for the algorithm as an example of an ideal athlete.


If you’ve read Jim Collins’ book, “Good to Great”, you know it is wise to have the least KPI’s possible. This helps people focus and be mindful of what is really important for the company.

When applied to a sales team, this analysis would show us if our KPI’s are working independently, working against each other or redundant. This would help us optimise our KPI’s, weeding out the ones that are redundant or less impactful and even generating better KPI’s that would replace others in numbers.

Key Takeaways

Data science gives us insight as to why our sales “athletes” are performing below their full potential and helps us help them with actionable and specific coaching. Furthermore, by using data science we can simplify and optimise our KPI set, putting out a clearer and more actionable message as to what the company’s goals are.

Originally published at on March 16, 2020.



João Cruz