Sam Allardyce recounted a humorous tale which re-enforced how important it is to have the right facts and figures at your disposal, and the importance of controls in establishing a trustworthy dataset.

During pre-season training Allardyce allowed a certain player who was not living near to the club to perform some training sessions at home – saving him a lengthy commute, and (theoretically at least) increasing his wellbeing.  To ensure that he was indeed training and to allow progress to be monitored, the player was provided with a GPS and heart rate monitor – to track his movement and the intensity of his efforts (a foundation for data-driven analysis). 

When the player returned for training sessions at the club, his statistics didn't come close to matching the very impressive readings from his GPS and heart rate monitors.  Whilst initially this was put down to a glitch, it was a pattern that continued throughout the pre-season.  When Allardyce confronted the player to find out what was going on with his home-based training regime, it turned out that the player had (somewhat creatively) strapped the monitors to his dog – and sent him out for a run around the local park.

This highlights the challenge of maintaining data integrity where humans are involved.  We make mistakes, we cut corners, like water we will find the path of least resistance to the outcome on which we are measured.  If that means creating a new customer called Mr. Mickey Mouse (rather than searching on an antiquated database to find his real name and address) in order to get a sales commission, or indeed strapping a GPS monitor to a dog, then chances are we will do it.

Another significant challenge for many organisations when trying to become more data driven is the need to get data into one place so that it can be analysed.  There are two parts to this problem:

  • Firstly, ensuring that analysis is being conducted with the whole picture in mind.  To a certain extent this mirrors the point made in the previous blog – namely that organisations need a model to explain what they are trying to understand and improve, and it is the model which dictates which data need to be integrated.
  • But even with a model in place, all the component parts need to come together.  It is little use if, having decomposed a problem into measurable components, you do not re-integrate the findings from those components to paint the whole picture.

As Peter King explains of British Cycling: "A while ago we adopted an approach picked up from NHS. Our equivalent of their patient-centric care is being athlete-centric in our analysis. Anyone with input to an athlete's success would regularly meet and talk through their needs and ideas. They would all put together their best view of what would be best for that particular athlete".

Through their athlete-centric approach which integrated all analysis, cycling was able to make significant changes which would not have been spotted had divisions remained discreet.  A good example of this: making sure that the cyclist and the bike are assessed as a single entity.  Seemingly obvious, but it had not always been the case:  "For a long time it was considered that narrowing the bike was the best way to streamline it. We spent a lot of time in wind tunnels over the years and probably do it better than most. But before the London 2012 Olympic Games we looked completely differently at how the front forks work with the wind. We realised that if you made them much wider they would actually create an airflow over the legs and body parts of the cyclists which makes the whole more aerodynamically efficient. Designed not to be efficient in their own right, but work perfectly when combined."

As such, cycling avoided silos of analysis, which curses so many businesses – creating a focus on the end goal which brought all components together.

But Dr. Marco Cardinale points out that not all coaches or clubs have managed to overcome this challenge: "[Data integration] is still a big challenge in the world of sport. Lots of people collect data in different places. In a typical football club, a sport scientist will have data generated by different data capture systems and formats and the medical team and performance analysts will be using separate systems and not integrating. The data is not in the same place and cannot be interrogated properly".

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.