Data Teams are like Role Playing Games (RPG). If you’re not familiar with RPGs, there is a person or group of characters all working together for a common goal. A crucial part of the individual characters are their levels, skills, and stats. In many games, higher levels are required to unlock specific skills. Likewise, stats show how well a character can utilize their skills.
In many RPGs, there are classes. These equate to core abilities that the character is good at. For example, there are soldiers, rogues, and mages. The soldiers are better at weapon fighting than a mage, whereas the mage is better with magic and casting spells than the soldier. The key to a cohesive group of characters is to have all of the classes complement each other. This complimentary fit allows each class within the group to focus and do its job well.
Data teams have three main classes: data scientist, data engineer, and operations engineer. You need all three of these classes to fight the big data battle. A common issue is to try to have a data team made up solely of data scientists. This need is like only having mages in your group. The data scientists get mowed down by the complexity of the distributed systems.
Sometimes RPGs introduce hybrid classes that try to combine two classes, such as a battlemage (a mix of soldier and mage). Likewise, the data teams have a hybrid person, a machine learning engineer that is a mix of data scientist and data engineer.
Levels, Skills, and Stats
As a data team starts out, each person will likewise have a set of levels, skills, and stats. Let’s take the coding skill as an example. The stats for the coding skill can vary significantly on a data team, from non-existent to mastery. Not every person on the team will have the coding skills, and even fewer can achieve mastery. It’s important to realize that some people won’t progress their coding skills from non-existent or from rudimentary to intermediate. This is akin to having your mage class be a high hit point tank, it just doesn’t work.
Data teams have a common problem: management takes a group of people with low coding skills and expects them to create a complex big data system. They think that taking enough people with low coding skills will average out to one person with mastery. It doesn’t work that way. It takes time and effort for individuals to progress (level up) themselves to a level where they can tackle complex distributed systems problems.
In RPGs, the characters face different challenges, usually in the form of some monster or beast. Big data and machine learning are formidable beasts if you didn’t know them already. When data teams come up against it without the proper levels, skills, stats, and classes, they will lose the battle or continually fight it in vain. The struggle will become so futile that your characters will quit (die). It’s critical that management evaluate if the team is ready for battle with the next-level monster.
Sometimes, teams will only try to fight easy monsters. Going up against easy monsters can make the team erroneously think they’re ready for much, much harder monsters. Data teams do this too. They fight the easy battles with technologies that make them more manageable and think they’re ready. The problem comes from teams with low levels and stats that don’t understand the layers of complexity involved in big data.
Evaluating and Fixing Your Group
Your data team won’t walk around with a number over their head showing the skills, and there won’t be a notification any time their skill increases, and they’re ready to level up. This growth is where the management of data teams comes in. Managers will need to honestly evaluate where the team is skill-wise and figure out how to increase it. Managers will need to figure out the size and scope of the monsters (projects) that the team can defeat. By leading effectively, the group will gradually level up to bigger and better teams capable of creating enormous value with data.
Check out my Data Teams book to learn more about leading your data team to victory.