Re: "Pursuing a career in analytics feels like pursuing a career in “science”—it’s not specific enough to go anywhere. Yes, there is a loose set of attributes that all scientists have; they are empirical, skeptical, observant, and structured."
As a scientist-turned-person-who-does-data-stuff, my empirical observation is that most people don't bother to even RTFM. I'll throw out a hand-waving made-up statistic/proclamation that reading the docs/performing a literature search in any domain - consistently - will put you in the top quintile. Some combination of grit, luck, mentorship, innate ability, and the like will determine the rest,
IMO, once you have strong fundamentals, any further ranking is going to be highly contextual at best, an exercise in false precision in most cases, and creates pathological incentives if taken too far.
I think my question for that is, which manuals do you read? Not like, going through one of those "data science curriculum" that's a long list of python packages, but just learn some of the basics of the math stuff and programming stuff and business stuff - and then what? I think I'm with you that that makes you "good enough" at all of them, so the next thing you should do is choose some business thing (or "data visualization" or whatever if you want that to be your thing.)
By "manuals" I mean the manufacturer's instructions in the case of an assay (for example in wet lab research), the basic theory for analysis of results (again e.g. for how to perform the calculations to go from raw to processed), or the vendor docs (e.g. any DBMS).
I don't put much weight on courses, not because the courses aren't good (many aren't but that's another topic) - but rather because most people treat them like a primary means to learn a subject quickly, which is magical thinking. An overwhelming majority of questions I see - and have seen over nearly two decades - are addressed by "the docs."
Edit: re the next thing you do: you build on the fundamentals and become increasingly useful, I suppose. I guess my point is I agree that "analytics" isn't a field as much as it is a means to an end.
Ah, you mean the literal docs. I thought you meant metaphorical docs, like trying to learn the foundations of stuff rather than just winging it. Which, yeah, people definitely do not read those. (And agreed on courses. I don't think those are particularly useful at all, but they do constitute a bit more of a formal training than just learning on the job. Whether or not that training is useful at all or not is a very different story.)
I find docs valuable for foundational learning. If you aren't familiar with the domain at all, they are a useful reference for expected behavior - a framework to use until you gain enough experience to deviate. If you are familiar, they provide insight into the philosophy of the tool/platform.
The idea of the very best being much, much better than the merely very good is the thing that has changed my thinking most this year
There's a david foster wallace quote that encapsulates this well. He was a nationally ranked junior tennis player as a teenager.
"The idea that there can be wholly distinct levels to competitive tennis — levels so distinct that what’s being played is in essence a whole different game — might seem to you weird and hyperbolic. I have played probably just enough tennis to understand that it’s true."
I wrote about this more here; I see this phenomenon everywhere now
"The idea that there can be wholly distinct levels to competitive tennis"
Every sport. Very obvious in pro golf, for example. There are amazing golfers out there who are probably 98th percentile and they can't make the PGA tour. Someone who can win consistently on the PGA tour is so next level it's indescribable.
Once in my life, I sat courtside at an NBA game, and, oh my god what they are playing is not basketball. It is 10 Terminators, blown up to video game proportions, throwing no look dimes while WWE'ing on a trampoline in a dead sprint. It was terrifying.
It reminds me of a chat I had with a friend who’s flown commercial aircraft for 30 years. At some point, how good of a pilot do you need to be? Your main job is to fly safely and not crash. This frustrates pilots because there’s an asymptote to “being a good pilot.”
I feel like that would be the ultimate dose of reality, where I realize this blog is a nice little recreational runner, and there are Olympians out there who would completely steamroll me in every conceivable way.
(Still, also, I guess I've thought about that a little bit? Though I have no idea what it would actually be about?)
Sometimes it's important to remember that comparaison is the thief of joy! So it comes down to why you would want to do it. You can go the easy way and maybe put together some of your ✨best✨ blog posts (I would buy even if I've read them already, just to support you). You could also even trace the shifts in (the) data (industry) by looking through your posts maybe. Or just write something completely new about data, careers, the future, life, etc. As you always do, easy peasy lemon squeezy🍋
Repackaging blog posts feels like such a cheat (though, I very much appreciate that you buy it anyway). But something completely different seems more fun. One day, I swear, I'm gonna write my giant magnum opus on Pitbull.
Great article, makes you think about what your 99% skill is.
One vocabulary I was missing is the glorified "T-shaped profile". The flat part would be the 95%, while the stem is the 99% (or 99.9%?). Having an abstract "Data" in the stem is arguably not the best strategy.
"What is your (potential) 99% skill" is an interesting question to me, because all of the examples people use for this (that I used, and other people used in comments) are sports, or academic talents with grades, and stuff like that. Which makes sense, because the gaps are so quantifiable and concrete. But it makes you wonder about the places where that is so much harder to see. Surely there is the same distribution in softer skills and other careers, but how does it surface? And how do you recognize when you're great at a thing where your greatness isn't measured in minutes and seconds? I really don't know.
How does a progress in skill level in Analytics (the stereotypical data thing) compare with other roles such as Software Engineering (some dataish roles like ML Engineer might be in here) or Product Management? I would argue that the technical core part of all these roles (the 95%) is relatively easy to master by just going through courses and doing small projects with friends (unless you are working on spaceships). The mastery comes from being able to integrate these skills into larger organisations in different industries (maybe that is the 99%? Could be called a soft skill, but maybe it is in fact the skill). This is not so much about knowing which buttons to press but more about how to communicate and align with other people. Unfortunately, this can only be trained by being part of a larger org, so there are some restrictions regarding training capacity. Maybe Analytics is in the most challenging position since it requires an org of staggering complexity in order to be justifiable.
hmm, that makes me think something different actually. Which is that, for the more technical roles, it can be useful (and potentially very valuable) to keep getting better at the technical stuff. Like, sure, engineering jobs are about soft skills and all of that too, but if you are a very very technically strong engineer, you can accomplish a lot more than a good but not great one, if you both have the same skills in other areas (eg, there's marginal benefit of additional technical mastery, regardless of your current technical mastery).
I'm not sure how true that is for data stuff. It's there, a bit, but there's probably much more of a "good enough" threshold in data, where additional technical mastery doesn't mean much.
This is exactly how I feel about my "career", like the wind, "everywhere but nowhere in particular". I'm the type that gets bored easily, so when my analytics job didn't seem to go anywhere, I thought it was because I wasn't interested in analytics as much as I thought. Interestingly, I have looked into economics, finance, and even trading as alternatives. But your post made me realize that it's not a CHOICE between analytics and other fields, but it's about the COMBINATION of both my analytical skills and domain expertise. I have a much better idea of what I could do next to grow my career now. Thank you for the insights, you've undoubtedly saved me years of agony in the future.
(jk jk, that's awesome, and I'm really glad it was helpful! And if you have any other epiphanies along the way, please let me know, so that I can tell my, uh, friend.)
Definitely agree that the difference between 95th and 99th is usually a massive gulf. In track (if we ignore PEDs), those 99%-ers are people with off the charts natural ability coupled with years of steady hard work and a fair amount of luck (it can be hard to avoid injury).
Yeah - I can't count the number of times I've seen someone run the WR and post-race they are running around like they didn't exert any effort. Must be adrenaline but my good sir/ma'am, you just ran faster than anyone ever!
I believe it! That's asking a lot for anyone in HS who hasn't trained. That's also a really odd requirement to play baseball. It's ~130 meters to run all 4 bases and how often does that happen? Better to train like a 100m sprinter if anything. It's a power sport.
I think it was mostly just a way to weed out kids who weren't serious, since you had to train at least a little bit beforehand. That, and hazing, basically.
Re: "Pursuing a career in analytics feels like pursuing a career in “science”—it’s not specific enough to go anywhere. Yes, there is a loose set of attributes that all scientists have; they are empirical, skeptical, observant, and structured."
As a scientist-turned-person-who-does-data-stuff, my empirical observation is that most people don't bother to even RTFM. I'll throw out a hand-waving made-up statistic/proclamation that reading the docs/performing a literature search in any domain - consistently - will put you in the top quintile. Some combination of grit, luck, mentorship, innate ability, and the like will determine the rest,
IMO, once you have strong fundamentals, any further ranking is going to be highly contextual at best, an exercise in false precision in most cases, and creates pathological incentives if taken too far.
I think my question for that is, which manuals do you read? Not like, going through one of those "data science curriculum" that's a long list of python packages, but just learn some of the basics of the math stuff and programming stuff and business stuff - and then what? I think I'm with you that that makes you "good enough" at all of them, so the next thing you should do is choose some business thing (or "data visualization" or whatever if you want that to be your thing.)
By "manuals" I mean the manufacturer's instructions in the case of an assay (for example in wet lab research), the basic theory for analysis of results (again e.g. for how to perform the calculations to go from raw to processed), or the vendor docs (e.g. any DBMS).
I don't put much weight on courses, not because the courses aren't good (many aren't but that's another topic) - but rather because most people treat them like a primary means to learn a subject quickly, which is magical thinking. An overwhelming majority of questions I see - and have seen over nearly two decades - are addressed by "the docs."
Edit: re the next thing you do: you build on the fundamentals and become increasingly useful, I suppose. I guess my point is I agree that "analytics" isn't a field as much as it is a means to an end.
Ah, you mean the literal docs. I thought you meant metaphorical docs, like trying to learn the foundations of stuff rather than just winging it. Which, yeah, people definitely do not read those. (And agreed on courses. I don't think those are particularly useful at all, but they do constitute a bit more of a formal training than just learning on the job. Whether or not that training is useful at all or not is a very different story.)
I find docs valuable for foundational learning. If you aren't familiar with the domain at all, they are a useful reference for expected behavior - a framework to use until you gain enough experience to deviate. If you are familiar, they provide insight into the philosophy of the tool/platform.
Every once in a while, I find myself reading them because I'm looking something up, and I'm always sort of amazed by how much I learn from it.
The idea of the very best being much, much better than the merely very good is the thing that has changed my thinking most this year
There's a david foster wallace quote that encapsulates this well. He was a nationally ranked junior tennis player as a teenager.
"The idea that there can be wholly distinct levels to competitive tennis — levels so distinct that what’s being played is in essence a whole different game — might seem to you weird and hyperbolic. I have played probably just enough tennis to understand that it’s true."
I wrote about this more here; I see this phenomenon everywhere now
https://residualthoughts.substack.com/p/its-lonely-at-the-top?r=9c2r
"The idea that there can be wholly distinct levels to competitive tennis"
Every sport. Very obvious in pro golf, for example. There are amazing golfers out there who are probably 98th percentile and they can't make the PGA tour. Someone who can win consistently on the PGA tour is so next level it's indescribable.
Yeah it’s counterintuitive to me!
Another example of this is higher ed. I have always thought of colleges in rankings, where #1 is a little better than #2 etc
But Harvard’s endowment is like 5x Cornell’s. It’s an entirely different level
Once in my life, I sat courtside at an NBA game, and, oh my god what they are playing is not basketball. It is 10 Terminators, blown up to video game proportions, throwing no look dimes while WWE'ing on a trampoline in a dead sprint. It was terrifying.
It reminds me of a chat I had with a friend who’s flown commercial aircraft for 30 years. At some point, how good of a pilot do you need to be? Your main job is to fly safely and not crash. This frustrates pilots because there’s an asymptote to “being a good pilot.”
Nah, I love it when my pilots do a little bit of acrobatics before they land.
Ben, when are you going to write a book?
I feel like that would be the ultimate dose of reality, where I realize this blog is a nice little recreational runner, and there are Olympians out there who would completely steamroll me in every conceivable way.
(Still, also, I guess I've thought about that a little bit? Though I have no idea what it would actually be about?)
Sometimes it's important to remember that comparaison is the thief of joy! So it comes down to why you would want to do it. You can go the easy way and maybe put together some of your ✨best✨ blog posts (I would buy even if I've read them already, just to support you). You could also even trace the shifts in (the) data (industry) by looking through your posts maybe. Or just write something completely new about data, careers, the future, life, etc. As you always do, easy peasy lemon squeezy🍋
Repackaging blog posts feels like such a cheat (though, I very much appreciate that you buy it anyway). But something completely different seems more fun. One day, I swear, I'm gonna write my giant magnum opus on Pitbull.
Great article, makes you think about what your 99% skill is.
One vocabulary I was missing is the glorified "T-shaped profile". The flat part would be the 95%, while the stem is the 99% (or 99.9%?). Having an abstract "Data" in the stem is arguably not the best strategy.
"What is your (potential) 99% skill" is an interesting question to me, because all of the examples people use for this (that I used, and other people used in comments) are sports, or academic talents with grades, and stuff like that. Which makes sense, because the gaps are so quantifiable and concrete. But it makes you wonder about the places where that is so much harder to see. Surely there is the same distribution in softer skills and other careers, but how does it surface? And how do you recognize when you're great at a thing where your greatness isn't measured in minutes and seconds? I really don't know.
How does a progress in skill level in Analytics (the stereotypical data thing) compare with other roles such as Software Engineering (some dataish roles like ML Engineer might be in here) or Product Management? I would argue that the technical core part of all these roles (the 95%) is relatively easy to master by just going through courses and doing small projects with friends (unless you are working on spaceships). The mastery comes from being able to integrate these skills into larger organisations in different industries (maybe that is the 99%? Could be called a soft skill, but maybe it is in fact the skill). This is not so much about knowing which buttons to press but more about how to communicate and align with other people. Unfortunately, this can only be trained by being part of a larger org, so there are some restrictions regarding training capacity. Maybe Analytics is in the most challenging position since it requires an org of staggering complexity in order to be justifiable.
hmm, that makes me think something different actually. Which is that, for the more technical roles, it can be useful (and potentially very valuable) to keep getting better at the technical stuff. Like, sure, engineering jobs are about soft skills and all of that too, but if you are a very very technically strong engineer, you can accomplish a lot more than a good but not great one, if you both have the same skills in other areas (eg, there's marginal benefit of additional technical mastery, regardless of your current technical mastery).
I'm not sure how true that is for data stuff. It's there, a bit, but there's probably much more of a "good enough" threshold in data, where additional technical mastery doesn't mean much.
This is exactly how I feel about my "career", like the wind, "everywhere but nowhere in particular". I'm the type that gets bored easily, so when my analytics job didn't seem to go anywhere, I thought it was because I wasn't interested in analytics as much as I thought. Interestingly, I have looked into economics, finance, and even trading as alternatives. But your post made me realize that it's not a CHOICE between analytics and other fields, but it's about the COMBINATION of both my analytical skills and domain expertise. I have a much better idea of what I could do next to grow my career now. Thank you for the insights, you've undoubtedly saved me years of agony in the future.
damn, I gotta start charging for this thing.
(jk jk, that's awesome, and I'm really glad it was helpful! And if you have any other epiphanies along the way, please let me know, so that I can tell my, uh, friend.)
Haha, sure will!
More than a little chuffed that you used an example from my favorite sport. That was a spectacular race.
Nguse and Ingebrigtsen also had a duel at the Prefontaine meet in September 2023 and I got to be trackside for that. 3rd and 4th fastest miles ever.
https://www.youtube.com/watch?v=evhyiluiEzI
Definitely agree that the difference between 95th and 99th is usually a massive gulf. In track (if we ignore PEDs), those 99%-ers are people with off the charts natural ability coupled with years of steady hard work and a fair amount of luck (it can be hard to avoid injury).
That shot of their faces as their starting the last lap - they don't even look the least bit phased. I do not understand.
Yeah - I can't count the number of times I've seen someone run the WR and post-race they are running around like they didn't exert any effort. Must be adrenaline but my good sir/ma'am, you just ran faster than anyone ever!
We had to run a mile in under 6:45 to make our high school baseball team, and honest to god half of us threw up afterwards.
I believe it! That's asking a lot for anyone in HS who hasn't trained. That's also a really odd requirement to play baseball. It's ~130 meters to run all 4 bases and how often does that happen? Better to train like a 100m sprinter if anything. It's a power sport.
I think it was mostly just a way to weed out kids who weren't serious, since you had to train at least a little bit beforehand. That, and hazing, basically.