Time for automation? It's less common than you think...
Last week I described five levels of formality that processes and components of your data systems can live at, ranging from Informal (Level 0) to Automated (Level 4). If you’re like me, your first inclination is probably to try and get everything to Level 4 as quickly as possible. I hear you. I see you. And I want to, as gently as possible, tell you to stop.
A lot of the mistakes I’ve learned from in my career have stemmed from trying to automate things that weren’t ready. So this week, in order to spare you the same hard lessons (if I’m not already too late), I want to explain some of the reasons you may want to put off or completely avoid going to level 4.
The investment isn’t always worth the payoff
Moving a process or component up a level or two costs time and energy. It’s an investment whose payoff is that the output is more consistent and, at least for levels 3 and 4, you save time each time you go through the process. The potential payoff depends on a lot of factors, particularly how often you’re going to run the process. If you do something once or twice a quarter, you’re probably fine at level 1 or 2. If you’re only ever going to do a process once, you may even be fine at level 0.
If the payoff isn’t worth the development cost, put that investment into a different component.
Sometimes you need the flexibility
Sure, those higher levels of formality provide consistency that will save you time and headaches down the road. But the lower levels have a big benefit too: Flexibility. If the nature of a process is evolving, or just fundamentally inconsistent, then raising its level of formality will just make the process break more often. Or you’ll have to keep investing more in updating it. Adding columns to your plate map template. Adding corner cases to your naming conventions. The reason the higher levels are more consistent is because they’re harder (and thus more expensive) to change.
If the fundamental process is evolving, wait until it settles down.
Adjacent processes matter
This one is easy to miss. In fact, this is the reason why thinking in terms of these levels is so important: It is very difficult, if not impossible, to upgrade processes to level 3 or 4 if adjacent processes are at 0, 1 or sometimes even 2. Let’s say you want to start automating analysis (Level 4) but the metadata is collected in inconsistent spreadsheet formats (Level 0 or 1). That means that before you run each analysis, someone needs to figure out what the person who made the spreadsheet was thinking, then put it into a consistent form. You can’t automate that. Maybe you can write some scripts to help them (Level 3) but more likely they’re going to be stuck at Level 2.
If adjacent processes are at a significantly lower level, upgrade those first, then come back to this one in a later iteration.
Defining processes takes time
OK, so let’s say you’ve got a component where you’ve weighed the cost and payoff, you know it’s consistent enough and all the adjacent processes are at high enough levels of formality. Before you jump in to upgrade it, there’s one more consideration: Do you understand what the process should look like?
Even if the nature of the process isn’t changing, it’s important to understand the practical details before you give up the flexibility that you have at lower levels of formality. If you can improve the process at this level, the payoff will be even bigger once you do upgrade. If you’re still figuring this out, it’s OK to wait a bit longer.
Conclusion
The point of all of this is that just because there are five levels doesn’t mean it’s a race to the top one. For a lot of components and processes, 2 or below is fine. For early stage biotechs, that applies to pretty much everything. In fact, that’s probably the best place to be. The point is to decide what level is right, then figure out how to get there.
And if you want some help figuring out the right level of formality for each part of your data system, sign up for a System Evaluation. I’ll walk you through a detailed rubric and create a detailed report identifying the parts of your data processes and infrastructure you need to address today, and what can wait for later. Sign up today!