A player summary from an English Premier League Club.

Dashboard Designer: Andy Kirk (visualisingdata.com)

Organization: An English Premier League club

Note the data in this example is genuine game data for an anonymous player in the English Premier League in the 2015-2016 season. Andy Singleton is not a real player name, and this data is not from a Liverpool v. Manchester United game.

Scenario

Big Picture

You play on an English Premier League soccer team. You played a game on Saturday and spent Sunday in recovery. It's now Monday, and you're due at the training field at 10 a.m. for a team debrief on the match. How did you do? Just as you arrive, you receive your personal player dashboard via email to your iPhone. It shows your physical performance metrics for the match:

• How did you perform in the most recent game?
• How did your performance compare to that of the most recent games and to the rest of the season?
• Did your performance match the rest of the team's?
• How was the team's performance relative to that of the rest of the season?

Specifics

• Players need to understand their match-day physical performance and see several key metrics, including total distance run, number of sprints, and top speed.
• Players need to know if their performance was above or below average. This helps them relate their performance to the coach's desired tactics for the specific match.
• To add context, players need to know how their individual match performance compared to their performance in the most recent games (especially, e.g., if they are recovering from injury). For deeper context, the entire season's games are shown too.
• Finally, each player needs to know how his performance, good or bad, was related to the team's performance on match day.
• Everyone receives the dashboard in time for arrival at training, encouraging data-driven conversations and even a little friendly rivalry.

Related Scenarios

• Any key performance indicator scenario could be applied here. The yellow dot would show the current state of the key performance indicator, the red ones could show recent periods, and the gray ones could show all time.
• You want to compare current and previous values for any discrete item (e.g., employees, products, countries, etc.).
• You need to track performance of individuals: How do their performance metrics compare to those of previous periods?
• You are in information technology, tracking the speed of your key overnight processes, and want an overview of most recent performances.

How People Use the Dashboard

In the first training session following a match for this English Premier League team, all players are required to arrive for breakfast at a given time. Just before arriving, they receive, on their phones, their personal copy of this dashboard. They look at the dashboard before they sit and chat with their teammates. As well as being a personal reference for their own performance, the dashboard also generates conversation at the right moment in the team's training cycle: when they are ready to review the most recent match. The timing of delivery is also engineered to be when the players get together: Management encourages competitive comparisons.

All players who played 70 minutes or more (in a 90-minute soccer match) receive the dashboard. If they didn't play the full 90 minutes, the metrics are normalized to a 90-minute scale. Metrics for players who played fewer than 70 minutes cannot be reliably scaled to 90 minutes; thus, they do not receive a dashboard for the match.

The dashboard shows each player how he performed across 11 key game metrics. The 11 metrics relate to running distances and speeds. Figure 11.1 shows the detail for the first metric, total distance run.

Great attention has been applied to the scales for the measures. They all display varying measures, with very different magnitudes. For example, total distance is typically many kilometers, but high-intensity run distance is less than one kilometer. All scales are drawn to a normal distribution, not a linear scale. A dot on the extreme right or left is a truly exceptional performance. Players have been taught to understand that the absolute position isn't the most important. Using a normal distribution means all the metrics can be compared with each other.

Data is only one part of an organization's decision-making structure, especially in professional sports. Players talk through the dashboards with their coaches. Coaches and players work to add context to the data. For example, a player might be recovering from injury. If that's the case, he can look at this match and the previous five to see if he is progressing as expected. Another example is if a player has a specific tactical objective in a given game. Perhaps he was tasked with marking a particularly fast and agile opponent. In that case, some of the player's statistics might be skewed above or below average, but in a way that was expected.

Why This Works

Good Use of Color

Color is used very cleverly in the dashboard, as shown in Figure 11.2. Yellow represents the most recent game and is used in the game name, date, dots representing most recent performance, and relevant stats on the right. Red represents the previous five matches and is deliberately slightly less vibrant than the yellow. The gray dots are visible but relegated somewhat into the background to reflect their relative role in this display. Green represents the team performance. The color palette is very basic but strong.

The dashboard does use red and green, which can be a problem for readers with color vision deficiency. On this dashboard, this problem is solved by double encoding all information as follows:

• The latest match is represented by a larger circle than the others.
• The previous five matches are midsize circles.
• The green “Team Rank” is differentiated by the column header and position.

Over time, as the dashboards become familiar, the colors enable fast recognition of the salient facts.

Scale Orientation

All measures are oriented so that dots on the right represent higher-than-average performance. As an example, Figure 11.3 shows top speed and recovery time. Top speed should be as high as possible, whereas recovery time should be as low as possible. In order to maintain consistency, all measures are oriented so that marks on the right are better.

Normalized Scales

The scales don't show a linear numerical scale. If they did, it would not be easy to compare relative performance of top speed and total distance, as one is a speed and the other is a distance. Once the scales are normalized, it is possible to compare the metrics in terms of their difference from an average performance. (See also Chapter 7.)

Minimal Use of Numbers

The only numbers shown are those for the most recent match. Other than that, the information is kept to a minimum so as not to overload players and coaches. Showing fewer numbers helps focus on the key goal for the dashboard: How did you perform in the last match, and how did that compare to other matches?

Social Interaction Promoted

Introducing data into soccer players' lives is a modern trend. The data is delivered directly to their smartphones at a time when all players come together for training. In this way, coaches know players have the dashboards, and the dashboards fit into a regular training schedule. This helps improve acceptance and usage.

The potential for players to compare their performances with others is deliberate. Competitive people love to have proof points showing where they did well. The timing and manner of the delivery of the dashboards to players is designed to encourage this.

Mobile Friendly

This dashboard is delivered to mobile devices and designed to be consumed on them. Its dimensions are optimized to a cellphone. It is not interactive. By having a static dashboard without interactivity, it does reduce the amount of information the dashboard can contain, but this is intentional. In a culture where data is being introduced slowly, a static, mobile, simple dashboard is a more prudent choice.

Quick Comparisons to Individual and Team

The prime purpose of the dashboard is to compare the player to his own recent performances. Figure 11.4 shows how this player's sprint statistics were below average in the recent match. However, what happens if it was a really bad match? What if the whole team played badly? Wouldn't that affect the player's results?

For this reason, the team rank, in green, on the right side of each metric, helps put a good or bad result into context. If both the player and team ranks are low, it indicates that the entire team had a bad week, suggesting that the result isn't as bad for the player as first thought.

The comparison to the player's other performances is the key objective of this dashboard. You might think that the comparison to the rest of the team is as important. Maybe that's right for some dashboards, but in this one, the objective is to benchmark the player against himself. The team rank is there not to let him see if he outperformed his teammates but if his personal performance might have been affected by overall team performance.

In soccer, comparing oneself too much to other players is not informative. Players in different positions have different roles and face different physical challenges and expectations. Additionally, perhaps a player had different tactical objectives that week.

Author Commentary

ANDY: A dashboard like this is not just about the data. It's about culture change. Many organizations choose to bring data into their culture slowly. Coaches' opinions about data in professional sports are divided. Some coaches think that sports such as soccer are so fluid it is not possible to quantify performances accurately. Harry Redknapp, one of the most successful English Premier League managers of all time, was famously antistatistics. After a losing game, he once said to his analyst, “I'll tell you what, next week, why don't we get your computer to play against their computer and see who wins?” (“How Computer Analysts Took Over at the Britain's Top Football Clubs,” The Guardian, March 9, 2014. https://www.theguardian.com/football/2014/mar/09/premier-league-football-clubs-computer-analysts-managers-data-winning)

Other clubs use data much more deeply, but nobody claims data should replace intuition and experience. The club is carefully introducing this example dashboard to help data inform players and coaches. It is a coaching aid, designed to inform discussion between players and coaching staff. Its simplicity, mobile-friendly design, and delivery timing all foster sharing and discussion of the data.

The simplicity of the design is intentional. One thing the analysts wanted to avoid was alienating players and coaches by providing too much too soon. Many extra things could have been added to this dashboard, such as:

• Interactivity, with tool tips for each dot, adding more context.
• Labeled dots showing actual values, not just position on a normalized scale.
• Buttons to switch from normalized to actual scale values to allow players to see more about their data.
• The result of the match.

Had the team introduced too much too soon, there was every chance the players and coaches would be overwhelmed with the volume of data. Instead, it's possible to educate players and coaches slowly and steadily, letting each new piece of information embed gradually.

Any organization bringing data into its culture for the first time could adopt this approach. Analysts often get excited about all the data they have and throw everything into their output in one go. Their unsuspecting colleagues in other parts of the business can become overwhelmed and, instead of embracing the new data-informed culture, retreat back into their Excel spreadsheets.

Remember: All dashboards have to compromise somewhere, often by omission. If you're starting off, it's better to start slowly and introduce complexity very gradually.

Jeff: We discuss the use of the traffic light colors in Part III: Succeeding in the Real World, and this is a good example of that color palette in action. This dashboard shows well in simulation for color vision deficiency because the dots in the dot plot are not red and green. The bright yellow dots show up very well in most cases, making it easy to tell the yellow dots from the red. Also, the dots are sized differently: The red dots are a little bit bigger than the gray dots, and the yellow are just a bit bigger than red. As seen in Figure 11.5, this helps distinguish the dots from each other. (For more on traffic light colors, see Chapter 33.)