Data scientists who strive for perfection often don't realize that managers are more impressed by employees who meet their deadlines. Here are ways to mitigate schedule slippage.
I've had to coach managers down from bouts of high anxiety because of schedule overruns with their analysts and data scientists. As a management consultant that works closely with data scientists and other IT pros, I know the challenges of staying on schedule, especially when engineers are involved.
A fundamental difference between the cultures of engineering and management is how they value time. Engineers value quality over time. That's not inherently bad; however, it does introduce a significant amount of schedule risk that must be mitigated.
What's most challenging (and often overlooked) for the analytic manager are the cultural aspects of execution excellence. Since this is a leadership issue and not a management issue, it's important that you and your change professionals take an active role in making sure your data scientists are hitting their dates. As a leader, it's important to build a culture of effective execution with your data scientists.
Let's look at some of the key reasons why data scientists tend to slip dates, starting with the most common one.
The path to perfection
There's the road to success and the path to perfection. With any feature, deliverable, or algorithm, your data scientists can probably get to an acceptable state pretty quickly if the plan is structured properly. This is where the road to success ends and the path to perfection begins.
These tiny little details that most of us overlook will incessantly bother your data scientists, and they will fight for perfection before calling it complete. I once worked with an engineer-minded manager who couldn't process the message on a slide if some of the bullet points ended in periods and some didn't. That's fine if you have the time to deal with these kinds of details, but most of us don't, so this perfection complex must be handled.
As a leader, the best way to mitigate the perfection complex is to support the idea that perfection is not required as long as we have success. Stopping at success is often difficult for data scientists, because they can't see where the path to perfection breaks off from the road to success. Work with your analytic manager to make this junction very clear. Furthermore, shift the responsibility of this decision from your data scientists to your governance team (e.g., business experts and testers) — they should decide what success looks like, not your engineers. Finally, don't let your engineers overrun success at the expense of the schedule — make it very clear that the initial focus is success, and we'll work on the details if we have time left over.
The second-guessing game
A second key reason why data scientists tend to slip dates is because they're confident they can second-guess the schedule without doing much damage. Data scientists are very bright people, so unfortunately they feel they can extend their brilliance into areas they shouldn't. One of these areas is the schedule, especially when they sense a date might be slipped. It's similar to the advanced calculus you undergo when the alarm goes off in the morning, and you need more sleep — suddenly Euclidean geometry comes in very handy.
So, they'll rationalize their slippage based on their understanding of the project schedule. I've had this happen to me on several occasions — I'll ask a team member about a slipped date, and they'll respond with, "Well, it's not on critical path, so it's okay." First, they don't know if it's on critical path because they never looked at a Critical Path Gantt; and second, it's not okay! A good analytic manager will put a lot of effort into a schedule — estimates, dependencies, schedule constraints, and resource leveling must be considered. There's no possible way even the brightest individual will know how much damage a date slip will cause based on a back of the napkin calculation.
To mitigate this, reinforce the idea that schedule development is not a simple task, and the dates in the schedule do not have an arbitrary tolerance. Bring the data scientists into the project management function and show them why the schedule is the way it is. It's important that the entire data science team — including the analytic manager — works as a team with the same objectives, rewards, and consequences. If a date is slipped, it's the whole team's responsibility to own the impact of that slippage, instead of throwing it over the wall for the manager to figure out.
A culture of execution excellence is a challenge with data scientists, but it's a requirement if you ever want your investment to bear fruit. Show them how to succeed without perfection and continually drive and support the value of staying on schedule.
I've described the two biggest reasons why date slippage occurs on data science efforts and key ways to mitigate them. Take some time to survey your team's execution capabilities, and work with your analytic manager on a cultural intervention if necessary. And, make sure you hit your own dates with this intervention — no schedule slippage allowed.