The “problem” with Goodhart’s Law is that we now know it exists. By “problem,” I mean using Goodhart’s Law when it comes to critiquing organizational metrics.
If you know Goodhart’s Law (rather, the rewording of it as we’ll see below), when you’re making metrics, you change them and adapt them over time before they get gamed.
When criticizing metrics (or anything, really) you should first assume that the people making them and using them are smart and trying their best…and know how to search the Internet. How could you help them make the metrics better, instead of shitting all over their attempts to do a good job?
Related: people in zombie movies seem to have never seen a zombie movie. If you know Goodhart’s Law, and you see a weird metric stumbling around on the street, you know not to go ask that metric if it’s in trouble. It’s a zombie!
And, as always, when you go and look at the original wording, the “law” isn’t really as all encompassing as we think, nor simple. The version we all use, according to Wikipedia comes from Marilyn Strathern in 1997: "When a measure becomes a target, it ceases to be a good measure." Hey! That’s good!
But, the actual one is: “Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.” I sort of don’t even know what that means, and I have a degree in philosophy. I assume monetary economist who get all excited about, like, interest rates and inflation know. My sense is that what the original is not saying is: “in a corporation, if you set KPIs/MBOs/OKRs and use those metrics to decide what decisions to make in the business, you will fail and are an idiot.” But that seems to be how we in the DevOps-y world reference Goodhart’s law. Like, it’s more about macroeconomic engineering at the country and global level? I think we can all agree that that system is super weird.
I mean, to apply one of my life principles (“good things are good, and bad things are bad”), I suggest a new “law” on metrics, in two parts: (1) shitty metrics are shitty (2) don’t use shitty metrics, use good ones.
We all worship at the alter of the DORA metrics. Like…METRICS. We’ve established those as an “observed statistical regularity” (did I get that right? Again, I couldn’t tell you if QE2 is economics or a new deodorant for 12 year olds), so does that make the DORA metrics bad? If not, why not? And if not, then, like, good metrics are good…?
My other suspicion is that a lot of anti-metrics talk is focusing on a symptom of a bigger problem: management at a company doesn’t really know what to do, what they’re doing, why they’re doing, how to engineer the rest of the organization doing it, and/or how to improve. They both lack clarity and are failing to make sure all the workers have clarity and, therefore, know what they should be doing.
In this case, whatever metrics you use have a higher chance of being bad. And as we know, bad things are bad.
Europe has entered a new age of anxiety – and it’s dragging Britain along too | Martin Kettle - “five overlapping big insecurities confront all Europeans. These are: the military threat from Russia; the stagnation and inequality of Europe’s economies; significant migration within and from outside Europe; the impact of climate crisis in remaking economic and social life; and the weakening of the nation state. Others could unquestionably be added to the list, not least the overmighty global power of the internet and of AI. And all of them connect.” // I’d make a slight edit: it’s a little early to tell if AI is going to tear apart society. Until it can play a Dungeons and Dragons DM perfectly, I think we’re pretty cool. And it’s pretty shit at that.
FinOps Open Cost and Usage Specification 1.0-preview Released to Demystify Cloud Billing Data - Trying to “establish a unified, serviceable framework for cloud billing data, increasing trust in the data and making it easier to understand the value of cloud spend.”
The Five Biggest Challenges in Digital Transformation - This is a pretty great talk from Laureen Knudsen on actually changing how your organization does software. It focuses on value stream mapping thinking and tools, but I’d simplify it to this: understand how your organization works, what it wastes time on, realize that you (“management” have the power to fix it and, like, it’s your job, and, so, go fix it. At the very beginning, there’s a great point: the good news it, the will to change is under your control, it’s not some external thing you can’t do anything about.
Cloud Native Users Struggle to Achieve Benefits, Report Says - Vendor commissioned survey, matches the anecdotes you hear: “Users are encountering issues around security, tool sprawl and cultural difficulties, including poor collaboration between developers, security and operations.” // At this point, I would chalk stuff up to (1) change is always hard, especially knowing and believing how much process change is needed to get the advantages of new tools (“culture,” as people like to poetically say), and, (2) expectations are always inflated: see the hype cycle. This is known.
Use Markdown in Google Docs, Slides, & Drawings - I didn’t know that was in there, must be new. I hear you can export to markdown too!
Amazon Q touted as the AI chat assistant for all things AWS - “A very small team of Amazon developers successfully upgraded 1,000 applications from Java 8 to Java 17 in just two days.” // Coworker: “Which team are you on,” sips coffee in the break room. You: “Oh, I’m on a very small team.”
GPTs for teachers - Whole bunch of prompts that look good and don’t seem to be hustling some slimy “workshop.”
There’s no half-wet blanket. The blanket is either wet, or it’s dry.
ChatGPTsplaining. Someone needs to explain to ChatGPT that it’s OK to just say “ok” most of the time and not overly explain every God damned thing.
“Read The Roomba” Here.
I’ve become part of the “time goes by in the blink of an eye” crowd. Like, I can’t believe the year is almost over! I feel like it hasn’t even started yet.