Breaking Down Advanced Data in Hudl Assist
Hudl is a leading international sports technology company that builds video analysis and stats tools for over 6.4 million pro and amateur coaches and athletes. Data/Analytics are growing in importance across the entire industry. Even small high school teams are exploring massive amounts of raw data and looking for new ways to turn it into important insights.
Hudl Assist takes the task of breaking down game footage off a coach's plate, so they can spend their time putting that data to work and help their teams win. Assist data helps teams understand their video and analyze their performance, make the scouting and game planning process more efficient and allows coaches to uncover trends that would otherwise take hours, or even days, to find.
The Problem: Tagging Subjective Data, Objectively
After Assist's beta run for the 2016-17 football season, user interviews showed that coaches wanted more advanced data in their breakdowns - offensive formation, defensive front, and blitz. These are tricky areas because there are hundreds of different ways to organize a team's players and coaches use a wide range of terminologies to describe these arrangements. We needed to figure out a way to define each of these columns of data, and how tag them quickly but with enough complexity to provide value to our users.
User Research: Coaches
I wanted to get a better understanding of how coaches tag formation, front and blitz data and how they use this data to analyze their game footage. Through a series of interviews, surveys and fulla queries, I discovered a few key takeaways. One, that coaches wanted accurate data above all else in order to be able to analyze their opponent's tendencies and predict how they will behave on the field. Two, that coaches wanted to have their games tagged in their own terms, but they were more than willing, even excited, to spend time naming more generic formations. The next step was exploring where in the workflow they would be able to access and name their formations.
At right: A sample survey question. This survey had a 27% response rate - woohoo!
User Research: Analysts
Once I understood how Hudl would define formation, front and blitz and how coaches use this data, I began to explore how our team of analyst could break down this data. One huge problem? Our analyst aren't football experts. We knew that they wouldn't be able to identify this information using standard football terms, let alone a coach's own naming conventions.
We decided to explore using diagramming as a means to an end as it’s more objective than naming and the analysts wouldn't need to remember formation names. I used another Hudl tool called Playbook to rapidly train analysts on diagramming and test whether they could identify and draw what they saw happening on the field.
This process proved to be successful; our analysts could accurately diagram formations. The next challenge was to make this process quick and efficient for analysts.
At left: A page from my quick and dirty training manual
Design Process
Team Sketch Session
No designer is an island, so I roped my team (product manager, devs, QA, the whole kit and kaboodle) into a sketch session. I set up the scenario, explained the assumptions, and gave a series of prompts which each team member would sketch a response to. We'd then present, hear another prompt, and iterate the design. This session was an incredible exercise is ideation, exploration and teamwork. In the end, it was a dev's pie-in-the-sky idea that we ended up shipping!
Everyone's a designer!
That Pie-In-The-Sky Idea
In the sketch session, one of the devs suggested we click directly on a video clip, which would place the players on a 2d diagram of the field. We all thought this idea was REALLY COOL, but was it feasible? In order the deliver a quality product, we needed to validate that it was even possible to tag formations in this way, and make sure our tagging and reporting mechanisms were simple and fast. With the goal of tagging every high school football game, a slight delay in tagging time or an issue with our backend could set off a chain reaction that could cause issues to ripple for weeks. Additionally, even if tagging games in this way was super fast, we needed to make sure our data quality didn't suffer for the sake of returning breakdowns to coaches with lightning speed.
Because this is already too long, here's a spoiler alert: it worked. And it was awesome.
Initial Mockups
Since I had a solid idea of the product's functionality from the start, I immediately started mocking up potential layouts for the workflow.
Just a normal amount of mockups.
Feedback and Iteration on the Offensive Formation Workflow
Through more training and testing, my team determined that drawing a formation would take less time than selecting one, so we placed emphasis on the drawing function by placing it directly below the video player. We also needed to give analysts the option to select a diagram from a list of commonly used formations.
Tagging a Game
To create a formation diagram, click on each player in the video at the moment of snap. The field diagram below will populate with each player as he is clicked.
Alternatively, a suggested diagram may be selected from the column on the right.
Clip Status
If an analyst has to leave the tagging interface at any time, he or she can easily navigate back to any untagged clips. In addition, clip status shows any issues such as untagged or problematic formations.
