• Infernal Ramblings
  • Posts
  • On Cohort Effects (or, Why I'm Starting a Newsletter About Working in Data)

On Cohort Effects (or, Why I'm Starting a Newsletter About Working in Data)

Introducing a newsletter for data scientists wrestling with the non-technical side of work (with a side of some thoughts on the internet and modern society)

Bottom line up front:

  • I’m John Lee, a data scientist well into my second decade of working in the tech industry and helping businesses make better decisions using their data

  • There seems to be a paucity of career resources for analysts and data scientists focused on the non-technical aspects of our role, and how we can grow our careers (you can find many blog posts on doing migrations and training ML models but hardly any on how analysts deal with executives who aren’t interested in looking at or using their work)

  • So I’ve started this newsletter hoping to fill this gap — and to also reflect on other things I’ve learned as I’ve grown in my career and left early adulthood

This newsletter will come out about every two weeks. A majority of content will cover thoughts from my career in data, but you can also expect reflections on the evolution of the internet and tech industry/startups, remote work, and navigating social and cultural differences.

For the first time in nearly two decades, I’ve got a blog again. Blog, newsletter, to me the two are nearly synonymous. As a Millennial on the older side in terms of my internet experience — my early usage of the web was shaped more by IRC, Geocities, forums, and blogging than by any sort of modern social media — it seems fitting that we’ve come full circle in the post-Facebook era and are back to reading blogs, just under a different name.

As I write this in February 2025, I’m coming up on the end of a 9-month sabbatical. My wife and I have been very fortunate that we were able to leave our jobs last April and just take a break from working. It helped, of course, that we’ve been working in tech in San Francisco for most of our careers, which helped us build up some savings. It’s helped even more that we chose to move to Kuala Lumpur, Malaysia, where I’m from, where the lower cost of living has literally helped stretch our savings farther.

What have I been taking a break from? I’ve spent a decade and a half working as a data scientist — or, what we just called an analyst when I started my career in the relative dark ages before “big data” or AI became the new hot things. For more than half my career I’ve been managing teams, and certainly all my career I’ve had to manage my time, my initiatives, and quite often larger projects even if I had no team formally reporting to me.

So if you ask my wife, this 9-month break has been 4 or 5 years overdue — it’s been coming ever since I had to lead a Growth Data Science team cut in half by covid layoffs, which then spent a year managing growth that had suddenly doubled overnight from pre-covid levels.

Don’t get me wrong: leading a team at Thumbtack, a home services platform, in the middle of covid was a fantastic professional experience — not least because after we got through the initial scare of covid driving our business to near-zero, we saw newfound consumer appreciation for how our technology could safely connect people to handymen and contractors they could trust. But I’d be lying if I said I didn’t still have some scar tissue from that year.

A Gap I Hope this Fills

Throughout my career, especially going through difficult times like covid, I’ve sought out readings and advice for how I could be a better analyst, and more importantly a better mentor and manager for my teams. I’ve found plenty of good counsel in writings aimed at engineers — I particularly enjoyed Will Larson’s An Elegant Puzzle: Systems of Engineering Management, particularly because how he approaches org structure and building out processes has a lot to bear on similar problems in data teams. There are a plethora of books aimed at product managers and project managers. And for the first-time manager leading a team in a knowledge business, designer Julie Zhuo’s The Making of a Manager is self-recommending.

But somehow, there aren’t as many resources geared to analysts. I use the term “analyst” intentionally here, because beyond data scientists there are also analysts who work in Operations teams and Finance teams; and many more analysts working on the technical side whose HR teams still haven’t caught up to the industry fad of labeling them as data scientists.

We’re not totally starved of good content here — Katie Bauer’s one of the few data leaders I’ve found who’s spent time reflecting on how to be a leader as a data practitioner. Benn Stancil, co-founder of Mode Analytics (RIP; of all the dashboarding/BI platforms I’ve used in my career, Mode’s was the most enjoyable to work in) famously writes weekly about his observations on how the field of data work is evolving. And Locally Optimistic has made probably the best organized collective effort I’ve seen yet of trying to be a useful, ongoing resource for analytics leaders.

But that’s kind of it — I exaggerate only a tad. I haven’t yet found a book aimed specifically at a hungry analyst eager for advice on how to accelerate their career. And to be fair, there may not be a viable market for such a book: something aimed at the most generic type of analyst is going to struggle to find advice that consistently resonates across the range of departments that the typical business-facing analyst might report into (Finance, Operations, Marketing, Product or Engineering). And while there are thousands of people working in the specific type of analysis where I’ve spent most my career — business and product analytics for technology companies — it wouldn’t shock me if the number of us working here is just too tiny to be meaningful to any publishing house.

I’d like to think I have stories and experiences enough to fill a book, even if there’s not many people who’d buy it. And taking a long break from work has only confirmed to me that this field I’ve spent my career in is one that I care about enough to want my thoughts and reflections about this work — working in data, working in tech, working in startups — to not just be idle ponderings in my own brain forever. So here we are.

Who is this For?

One of my favorite data phenomena which I’ve reliably seen at every company I’ve worked at, across every industry, has been the cohort effect. With high confidence, we can predict how a population will behave over time, simply by tracking their cohort. The simplest example of this is mortality rates: take the cohort of all the people who were born the year I was, 1990. A predictable fraction of us are going to die every year. And, barring something miraculous, we can be confident that at some point in the future — say, by 2140 — none of us who were born that year will still be walking this earth. We know this because we can extrapolate mortality rates from all the cohorts that came before mine: the share of people born in 1990 who’ll survive to their 40th birthday is probably going to be pretty similar to the share of people born in 1984 who made it to their 40th birthday last year.

This may seem obvious, but knowing that future cohorts will behave in the same way as past ones helps us predict the future: even before any of us were born, we could be sure that when my cohort was 20 years old, we wouldn’t have any savings or meaningful disposable income on average. After all, that’s been true of basically every cohort of 20-year-olds in history. But today, in the middle of our careers, my cohort is making much more on average than we were 15 years ago. While far from all of us are parents or homeowners (I am still neither), many of us are, and this number will keep growing every year.

Actuaries could have quite accurately predicted all this — the fraction of us that would now be parents and homeowners — fifteen years ago, when most of us were still students or just starting our careers. They can confidently predict that by 2055, much of my cohort will be retired — barring, again, some extraordinary event. If you are a businessperson looking to sell products to young professionals, parents, or retirees, it behooves you to understand what’s happening to the cohorts of people who will become your future customers — and when your current cohort of customers will naturally age out of using your product.

In my last job at Pinterest leading the Business Operations & Strategy team focused on Growth and Engagement, one of my biggest responsibilities was forecasting user growth and explaining this forecast to the CEO and CFO. While there are a variety of techniques analysts use in forecasting, and different business questions we have to understand (how did a recent product launch affect growth; how much market share might we be losing or gaining from competitors), it often felt like all I and my team were doing was trying to understand the different cohorts of people using Pinterest, and what past cohorts’ behavior implied about how future cohorts of people would use Pinterest. So I hope you’ll pardon the by-now extended digression about cohort effects, because to me, this is what makes it worthwhile to write and share my thoughts about data and analytics and working in tech.

The breakneck speed at which things change in tech and in business makes it easy to forget that due to cohort effects, so many things in life are immutably predictable. As a child and young professional, I found it easy to dismiss how much my parents and older people cared about their health — what supplements to take, what foods to eat and what to avoid. And now as people in their 30s, my wife and I are of course on a cocktail of daily supplements and carefully planning our menu for the week based on what our doctors would say. No matter how much I might have laughed at the idea of becoming like my parents a decade or two ago, at least in this respect, it was inevitable. Cohort effects don’t predict everything for everyone, but for the average person, they do predict a lot.

And so I write in the hopes that my experience will still be relevant for the cohorts of people who come after me. The more you resemble me, reader, the more relevant my stories and advice presumably will be. Now, as a Malaysian who went to university in the US, worked in business and product analytics for almost 15 years, and spent my young and early middle adulthood in Washington, DC and San Francisco before moving with my San Franciscan wife back to Malaysia, I am quite confident that there are few, if any, people in the world who’re following my exact life trajectory. This cohort, if it exists, must be minuscule. But I think there are a lot of people out there who fit some part of this bill, and for you, dear reader, I hope this newsletter can offer useful reflections wherever our lives’ journeys and experiences may overlap.

Topics We’ll Cover

A lot of times in my career as an analyst and people manager, I’ve struggled with questions where I was certain I couldn’t be the only one to have faced this before:

  • how to deal with a manager insisting we use a project management system meant more for engineering than data work or analysis

  • how to prioritize seemingly equally and highly important and urgent data requests

  • how to manage disgruntled or disengaged stakeholders who have a vested interest in not using or understanding a piece of analysis

Of course, virtually everyone I know who’s worked in data long enough has run into these kinds of problems. And everyone has a slightly different approach to these things. But it’s struck me that there are so few of us who’ve written anything public-facing about how we’ve dealt with these in real scenarios. It’s possible to find CTOs, like Will Larson, writing about how they set an engineering strategy for their companies and teams. But rare is it to find a data leader reflecting about how they set a strategy for their team, or post-morteming what happened after the fact.

There’s a plethora of materials on the technical aspects of data strategy — how to decide on migrating from one data platform to another, or how to design and analyze an A/B test. But in terms of the day-to-day knowledge work as an analyst — and especially a leader of analysts — there’s precious little on the human side of working in an organization that’s counting on you to tell them what data suggests they should do. There’s little an aspiring data scientist can read about how to motivate their team, organize a project requiring help and input from multiple people, or convince a stakeholder to pay attention to their work.

So for my fellow cohorts of analysts and data scientists, I hope there are things I can share from my own experiences in the more human realm of data work that might be helpful to you. I can’t guarantee that my approach or my opinions will be right, or work for you, but I hope my reflections can prompt helpful reflection if you’re going through something similar to a situation I’ve faced before.

And even if you don’t work in data, but someone you work with or care about does, I hope I can be helpful. There’s not very much literature out there about working with analysts — seriously, just try searching “how to work with engineers” and compare those search results with “how to work with analysts” in your favorite search engine. (You can also try “how to work with data scientists” and the results won’t be any better.) To the extent that this literature it exists, it’s often a chapter in a book written by a product manager covering various aspects of “PM 101.” Even if they aren’t exactly my cohort, I hope some of my experiences can be useful for folks who work with or care about data people.

What’s Next

The last time I visited San Francisco, I caught up with a Malaysian software engineer friend in his 20s, who reminded me very much of myself a decade ago: missing home, hoping to move back to Kuala Lumpur one day, but very much unsure how this could happen amidst him putting down roots in San Francisco across his professional and social lives. He made sure to catch me while I was in town because there just weren’t very many people he could talk to about making the kind of transition I have, leaving a tech job in San Francisco to move back home to Asia and be closer to family.

While a lot’s changed about both Malaysia and San Francisco since I was his age, and in many ways we are not quite of the same cohort (his career opportunities as an engineer look quite different than mine as an analytical data scientist) a lot of questions he faced were ones I remember grappling with too: what kind of job would I be able to get? Would I actually enjoy being back home?

So I hope this newsletter will prove useful for many cohorts of people out there — be they people on a trajectory like or adjacent to my own, or just people curious about the often untold perspective of the analyst pulling the data that companies use to make decisions. While I plan to write predominantly about my career experiences in data science and analytics, I’ll also have personal reflections from time to time on adjacent things that I think have shaped my journey. A sampling of topics beyond data that you can expect from me:

  • The evolution of the internet and social media; as someone who grew up on the internet and has been socializing on it my whole conscious life, I find this fascinating, much more so than you’d think even from noting that the last two companies I’ve worked at are Pinterest and Cameo (who in many ways sit at completely opposite ends of the spectrum of social media platforms)

  • Remote work; as a remote worker and previously manager of remote teams, I can’t help but have a lot to say about this experience (especially now that I’m based in a timezone that’s about as far as one can be from Silicon Valley’s)

  • Life as a migrant or expat; being a migrant in both directions — after spending my 20s assimilating into American society, now in my mid-30s I have to figure out life in the country I grew up in but have never lived in as an adult — this can’t help but occupy a hefty chunk of my mindshare these days

I’m planning to write 1-3 posts per month, so you can expect something from me approximately biweekly (or fortnightly, as the Commonwealth would say). I welcome thoughts and feedback, so I plan to leave comments open on my posts. I’m also happy to hear from you directly on the social media platform of your choice, or my email. Just don’t forget to subscribe!

Reply

or to participate.