When should you automate?
"Everything Starts Out Looking Like a Toy" (No.38)
This week’s toy: a bracket maker … for anything. Why stop at basketball when you can create rankings for anything? If you want to read more about finding the local maxima of anything, read up on the Stable Marriage Problem of math. Edition No. 38 of this newsletter is here - it’s March 20, 2020.
The Big Idea
At What Point Should You Automate?
As a team member in a fast-growing startup, there’s often tension between working faster and working better / smarter / more effectively. If the task requires more effort to explain than it does to do, it feels like the trade-off is not effective. If the task takes a long time to do, and automating it feels like freedom, the need to automate feels overwhelming.
Of course, life doesn’t imitate expectations. XKCD pictures it this way:
Building a complex automation might create a new job that takes additional time and resource, without solving the underlying root cause. But doing nothing might create chaos, as the people involved in the original task scale … like people, not computers.
Threading the Needle
There are probably some elegant algorithms to determine what to automate and when, and I think of it this way. What are some tasks that given sufficient structure could be done by someone (or a process) with high certainty and accuracy? If your process doesn’t seem like it could be automated, write it down. Pretend you’re a robot.
Using an Outline as a Standard Operating Procedure is a good start. Write the main guideposts as section headings. List the outcomes you want to receive. Let’s say you needed to make sure that inbound leads in a sales system were standardized to have a first name and last name that begin with a capitalized letter and had valid a email address.
You could look at every lead as it comes into your system, and maybe that made sense at the beginning when you had a few leads daily. By the time you hit 50-100 in your system, that 2-4 minute process of checking the name and email address takes up a bit of time. Perhaps you ask someone to do the work for you, but they run into the same issue, and as you scale (and especially on the days when you get spam email bursts) that doesn’t work. Now you’re stuck. You need to automate.
So how would you go about automating a process like this to save yourself time? It could be as simple as imagining yourself following the data as it comes into your system.
First, someone comes to the website
Then, they fill out a form
The form has three to four fields {name, company, email, opt-in preference}
When they enter their email, check (in real-time) that the email address is valid; or that it does not come from a list of disposable email domains.
On insert to the system, check that the first letter of the first and last name is capitalized.
You might not know how to do all of the things in your process, and it might not be smooth on the first iteration. And building the map of how things should happen will help you build the blueprint for letting you deal with the exceptions to your process rather than every record. Don’t forget to define the outcomes. Setting up a process with no defined goal is also a recipe for sadness.
What’s the takeaway? Slow down to speed up. By examining the parts of your process that could be completed by a well-documented outline, you will get closer to a series of steps that are automated or completed by other resources.
A Thread from This Week
Twitter is an amazing source of long-form writing, and it’s easy to miss the threads people are talking about.
This week’s thread: on the bond yield curve. What is it, and why should we care?
Links for Reading and Sharing
These are links that caught my eye.
1/ “Enhance, Mr. Data.” - You’ve seen it in too many movies. The protagonist is viewing a satellite photo of something, and they want to see more detail. “Enhance 200%”, they say to the computer. It looks like Adobe has built just that. With super resolution, we’ve entered another uncanny valley of not knowing whether an image is captured/real or enriched.
2/ Pixar Mail Trucks - Have you wondered why the new USPS trucks are so weird looking? It looks like a result of many design studies, or perhaps someone has an odd sense of humor. Putting aside this Pixar-ish resemblance for a moment, what would a better “mail truck” look like? It would likely be electric-powered, easy to get in and out of, have a high visibility, and a high cargo capacity relative to its size and weight. UPS cargo bikes (see below) may be a happy medium.
3/ Time expands to fill the task - Parkinson’s law – work expands to fill the time allotted – is inescapable. So what can we do to use our time better? Aliza Rosenfelder writes about strategies to improve our time use. Note that this is not just a “work smarter, not harder problem” but one that has multiple vectors including time, cost, energy, and mental focus. There’s no one right answer and thinking about the problem can improve the way we get things done.
On the Reading/Watching List
The Falcon and The Winter Soldier is Peak MCU. By this I mean it delivers comic-book thrills (a great chase scene, some fight scenes, and character reveals) along with pathos, humor, and scale. I’m really looking forward to this as it continues revealing. (Disney+, weekly)
Grantchester is everything Disney+ is not. Slow, contemplative, life at a human scale, and British. This is a funny and sometimes serious detective show set in 1950s Britain with an unlikely buddy duo - a vicar and an inspector. (PBS and Amazon Prime, 5 seasons)
What to do next
Hit reply if you’ve got links to share, data stories, or want to say hello.
I’m grateful you read this far. Thank you. If you found this useful, consider sharing with a friend.
Want more essays? Read on Data Operations or other writings at gregmeyer.com.
The next big thing always starts out being dismissed as a “toy.” - Chris Dixon