The Narcissism of Small Differences in Charts
If you see a ranking chart where the difference between the top choice and the bottom choice is 10% to 15%, tread carefully.
The Narcissism of Small Differences in Charts
In surveys, just because something is last doesn’t mean it’s bad. It’s true the other way as well. Now, if there’s a huge difference between the first and last thing, then, sure. If 85% of people like the first choice, and 15% like the last, you’ve got a stew goin'.
But, when the differences are small - let’s say all below 50%, you need to be careful with how you interpret the chart.
For example, there could be six to ten technology choices. 43% of people think the top is cool (or are planning on using it…whatever “good” is). 40% for the next, 38% for the next, and so on. And then 30% of people like the last one.
To me, the difference between 43% (or 38% for the third one!) and 30% is likely a narcissism of small differences. If 38% of people would prefer to eat beef hamburger (or use MySQL) over 30% who’d prefer a breaded chicken sandwich (or use Postgres) - who cares?
What’s more interesting is (1) 3–5 year growth/decline rates (that’s the most valuable part, for example, of the years of the RedMonk Programming Language Rankings), and, (2) thinking about how few people choose the top choice (or bottom). For example, in this survey (chart above), 44% of people think open source is “proven,” with 52% being “passive,” and 4% saying it’s emergent. 4% is a valuable number because you can say, “look, everyone agrees that open source is good, at least we know they don’t think it’s shit.” Also, the chart above is an example of the opposite case to the narcism of small differences: there’s a dramatic difference between the top two items and the bottom three. The difference in “proven” between open source (44%) and the bottom choice, “low code” (8%), is very meaningful: developers do not like low code.
Looking at the top, I think only 44% thing open source is “proven”? That is crazy. Open source has been proven since the 90s, let’s say 1997 when Oracle database ran on Linux for the first time. Open source is beyond proven: it’s the way. It’s closed source that has to prove itself now, at least for new projects. (As ever, SaaS/cloud is a clever way to escape the the difficulty of selling free software!)
Now, I didn’t read that survey super close, so don’t stress about my interpretation there. I’m illustrating the point: don’t get caught in a chart’s narcissism of small differences. If there’s a huge difference between top and bottom (and the other categories), sure: you can do something with that. But if they’re all in the same range, focus on rates of change and how close the top choice is to universal, or not.
Here’s an example of focusing on the rates of change from our State of Kubernetes survey:
If you look at the bottom, they’re kind of in the same range. But, look at “Shortened software development cycles”: it’s been going down over three years. It’s only a drop of 14%, but it’s going the wrong way. Kubernetes is supposed to make things better, but as I commented on at the time, for developers this survey shows it making things worse.
If you see a ranking chart where the difference between the top choice and the bottom choice is 10% to 15%, tread carefully to avoid succumbing to the narcissism of small differences.
(I could have found much better example charts, but, believe me: they exist!)
This is my daily-ish newsletter with stuff like the above, commentary of computer industry stuff, and rando/weird/fun things I find throughout the day. Subscribe for free to get more. People say it’s “gold” and I just wonder why they don’t say it’s “platinum,” but I’ll take it. Well, anyhow: subscribe!
Developers! Run Your Enterprise Java and .Net apps in the Cloud
VMware and Microsoft partner on a fully managed and hosted suite of Spring, Java, and .Net services, running Spring Cloud and several (pardon the phrase, but, you know) secure software supply chain and developer tools that are fully integrated with it. It works! It’s great! It delivers on the self-service, #noops dreams we assumed cloud has had all along:
I don’t know, maybe that’s over-stating it. I stopped programming in 2005, so what do I know? Nevertheless, it’s good stuff.
We have two in-depth sessions on Azure Spring Apps tomorrow that you should check out if you’re a developer, architect, or whatnot.
I don’t know the co-host from Microsoft - I’m sure they’re fantastic! But I know Adib, and he’s a great speaker when it comes to explaining. (Also, check out his book on cloud native app security. You can get it for free.)
I say this frequently, but I’ll say it again: people need to stop being fancy and posting .svgs and .webm images on the web. NO ONE GIVES A SHIT. Post a jpeg or a png and focus on more important things.
“newsletter is god”
I love having something I need to do that I can procrastinate with so I can hold off doing something I need to do that I don’t want to do.
I don’t think I’ll ever get comfortable with those survey readout PDFs that are formatted in landscape rather than portrait. They’re reports, not presentations. If some chart doesn’t fit the width, just make a page that’s landscape all of the sudden that’s the big-ass chart.
Sometimes I need a MattLevineGPT that summarizes his newsletters.
I always remember things when you remind me.
Relevant to your interests
VMware Revamps Partner Program for Multi-Cloud Future - new stuff in the partner program.
OpenAI: ChatGPT Could Disrupt 19% of US Jobs, Is Yours on the List? - Researchers estimate that ChatGPT and similar software tools could impact at least 50% of the tasks needed for around 19% of US jobs. 80% of the US workforce could see at least 10% of their work tasks affected by ChatGPT, which recently upgraded to a new GPT-4 model.
OpenAI Shut Down ChatGPT to Fix Bug Exposing User Chat Title History - As of March 22nd, 2023 at 10:05am Amsterdam time, it’s still down for me. I pay for ChatGPT+ and rely on it, including the history. This is one of the longest SaaS downtimes I’ve ever experienced. // “OpenAI temporarily shut down its popular ChatGPT service on Monday morning after receiving reports of a bug that allowed some users to see the titles of other users’ chat histories.”
I have no idea if my videos are worthwhile
I did a lot of videos over the past few years and they were well received. I think. I haven’t done videos (aside from podcasts and interviews) in a while.
I can’t judge if the format is better/the same for me writing things. The leads and funnel stuff don’t really exist for videos and the view count is…ok for such a niche category, but no where near “great.” Well, that’s only sort of true: the view count in LinkedIn and Twitter is usually huge. But, the % time viewed get to about 70% at most, but only for the 60 second ones. On the positive side, people say they like them - but mostly VMware people…so are these videos more valuable to my career inside VMware, or actually useful for driving brand and thought-leadership for VMware leading to, of course, selling stuff? FUNNEL SHIT.
It is impossible to know!
It takes about the same amount of time to make a 60 second video as to write a good article/blog post. (Again, the podcast/interviews are a whole other thing.) This is an annoying, becoming more part of the problem than the solution thing, but it might be a good idea to just redo many of them as new videos. I see the numbers, and it’s for sure that not everyone in our market has seen them :)
Is the writing any better? Probably…?
Tomorrow I’ve got those two presentations that are acting as a mental blocker for my usual flow of work. They’ll be fine. Then I can get to two or three writing projects I’ve got queued up.
So far, one person signed up for the full on info dump newsletter - fun! If you want to try it, here’s a 15 day trial. I need to rename it to $cached_out. Puns! We’ll see if that works in substack. I’m thinking I’ll try to send this newsletter out in the morning, and the $cached_out on the afternoon. All the stuff I cut from this “more polished” newsletter can show up in the dump one.