The 2024 Commoncog Recap

At the end (or start) of each year I write a recap essay (2023, 2022), meant to summarise all that Commoncog has covered over the past year. Commoncog’s current pitch is that we help you get better at business faster. But the core offering and positioning has evolved significantly over the past two years; high-level recaps are sometimes the easiest way to track the progress we’ve made.

I’ll do a thematic retrospective first, before letting you behind the scenes to talk about what I’ve learnt (which is a nice way of saying “tell you about all the mistakes I’ve made in the past year”). If you’re new to Commoncog, reading this recap is a good way to catch up on everything that we’ve published in 2024.

Here we go.

Commoncog’s 2024, Recapped

The easiest way to think about Commoncog’s yearly output is that at any given time there are a couple of threads that we’re pulling on, and in parallel.

1. Some threads take time because they are basically field reports from practice — I’m testing some set of business ideas against reality (either in my business or in the business of friends), and it takes time to write up the results. Some of these experiments consume time with no writing payoff — if it fails, I tend not to write about it. A handful of these experiments may span years — for instance, we currently have three ongoing that I’ve never even mentioned on the site.
2. Other threads take time because they are case research projects. We decide on a particular direction, and then commission cases for the Commoncog Case Library.
3. Finally, there are side quests — random updates to ideas we’ve covered in the past, written opportunistically when an opening presents itself.

The background here is that Commoncog is slowly turning into a case-first publication. This means less ‘field reports from practice’ and more ‘case-directed research projects’. I first wrote about this in Commoncog’s Next Phase; the short explanation is that there is a strong science-backed reason to invest in cases as an expertise acceleration mechanism for business, but also that the field report model is flawed: as I get better at execution, the business experiments that I run are going to take longer to execute and the field reports are going to become harder to write. As an example, I applied April Dunford’s book on repositioning to a friend’s SaaS company in 2021 and doubled their annual recurring revenue in eight months with no additional marketing spend — a seven-figure increase in ARR. Four years later, I think it’s finally safe to write about. I may or may not write about it this year.

That said, this transition to a case-first publication is still in progress. I’ll touch on how things are going at the end of this recap.

For now, here are the threads that we’ve pulled over the course of last year:

The Data Driven Series

The Becoming Data Driven in Business Series is more than two years old at this point. It is also the series that, of everything that we’ve published in 2024, probably made the biggest impact on the world.

The two standout articles were:

1. Becoming Data Driven, From First Principles. In the opening of this piece I wrote “If I do this right, you will never look at data the same way again.” Judging by the response — numerous executives who have assigned this to their teams; various business owners who have transformed their businesses as a result of these ideas; operators completely changing the way they run things in their business units — I think the essay succeeded.
2. The Amazon Weekly Business Review. This is the best description of the Amazon-style WBR currently published (at least until early Amazonian execs Colin Bryar and Bill Carr release their materials!) It took me over a year to write, and in the end I’m glad it took that long: the piece is written from practice, and not just a regurgitation of what I learnt from Bryar. I like to joke that I now belong to the group of insane people who have implemented an Amazon-style WBR from scratch. This knowledge has been genuinely life-changing (I am eternally grateful to Colin Bryar), but has been rather hard-won.

There were a few other articles in the series that I’m proud of. For instance, Data is Just an Added Sense came from a conversation I had with the founder of a late-stage venture-funded company, who wanted to put the WBR to practice in his rapidly-growing business, but felt some trepidation (as well he should!). ‘Data is just an added sense’ is a solution to his worries, and captures an attitude towards data that I’ve observed to be present in every good, data-driven operator I know. I’m pleased to have finally figured out a way to articulate it.

The Limits of Operational Excellence talks about why operational excellence (along with the rigorous use of data) only matters for certain types of companies. The basic insight is from Brian Lui, and was something that I discussed briefly on my appearance on the Analytics Engineering Podcast: companies that benefit the most from operational excellence are the ones that earn a return on invested capital only slightly above the cost of their capital. We demonstrate this with a case study on Danaher.

People think that ‘Continuous Improvement’ is about improving continuously. It isn’t, not really. The Secret at the Heart of Continuous Improvement talks about something that most mainstream accounts of Continuous Improvement or Lean don’t seem to get: the secret that makes CI work is that it actually introduces a rigorous approach to single-subject (N=1) studies. This is a bigger deal than first meets the eye — it means you don’t have to do two-group, randomised-controlled trials to gain knowledge.

Finally, Making Sense of Deming was the hardest piece in this series to write. I concluded my two year deep-dive into statistical process control pioneer W. Edwards Deming, and summarised the entirety of the man’s work in one essay. The goal? To ensure that you don’t have to read the man’s work if you don’t want to. This is, I think, important: Deming is the single most credible thinker on the Operations side of the Business Expertise triad.

There is one final essay in the Becoming Data Driven series (on how to start doing an Amazon WBR), and it’ll be out in Q1 2025 if everything goes well.

Launching Xmrit

Something that amplified the impact of the Data Driven Series was the release of Xmrit (pronounced xa-me-rit) — a free tool to create XmR charts. For the uninitiated, an XmR chart allows you to differentiate between routine variation (changes in your metrics that you can ignore) vs exceptional variation (a change in your metric you should investigate). This ability to separate signal from noise unlocks one’s ability to become data driven. Process behaviour charts — of which the XmR chart is one — is deeply tied to Deming: his mentor, Walter Shewhart, invented the chart in 1924, and Deming popularised it during the Allied production ramp up in World War 2, and again in Japan during the post war period. The technique turned 100 years old in 2024. While it is not perfect, it gets you a long way towards becoming data driven.

We built Xmrit for internal use first, to bootstrap our feel for the variation in our metrics. But it quickly became clear that the tool was useful more broadly. We learnt that the vast majority of XmR chart implementations over the past 100 years were by folks in manufacturing or in healthcare, which meant that most of the available software wasn’t built for intuitive use. This was an opportunity. As software people, we were both willing and able to spend the resources to make XmR charts easy to create and share. I think we’ve largely succeeded.

We released Xmrit for free in April and made it open source shortly after. We also built and released the Metrics Masterclass — a paid course that teaches you how to use XmR charts in a variety of situations, compressing four books, a few dozen statistical papers, and months of painful trial and error into one self-paced video course.

Xmrit’s funding model is simple: the course pays for continued maintenance and development of the tool. The tool itself is free and open source, and we want it to be copied and adopted by as many other businesses and integrated into as many other tools as possible. (A fork of Xmrit’s code lives in our custom WBR software, for instance). If you’re reading this: please, take the code! Our contributions here are really the user interface innovations; the important thing is to copy the interaction model that we’ve uncovered.

The feedback has been meaningful: I’ve had many conversations with operators who realise that this was the thing that prevented them from using metrics seriously. Amongst other things, I taught a full-day workshop on XmR charts at Shopback late last year, as well as introduced Xmrit in a Reforge module at the invitation of Crystal Widjaja. Crystal has deeply influenced my thinking about data; more on this in a bit.

I want to give a special shoutout to Sam Taylor, my partner in Xmrit. Sam comes from a manufacturing background and was trained in Six Sigma. He was intrigued by the idea that XmR charts might be applicable outside of manufacturing. (To be precise, I presented him with statistician Donald Wheeler’s arguments, but Sam was truly only convinced after seeing it in practice in Commoncog’s WBR). Sam’s ability to reason about data, to check my tendency to jump to conclusions, and to bring up counter-factuals when sensemaking has proven to be invaluable over the past year. And I’d like to plug his writing too: the Xmrit articles he’s written about using data in business are some of the finest you’ll get on the topic.

I’d also like to thank Aw Hang Bin, Bennett Clement and Sean Lim for contributions to Xmrit. Hang Bin wrote the original version of Xmrit as a computer science freshman; Bennett wrote and rewrote the entire codebase multiple times in the months afterwards, and Sean migrated the tool to Apache ECharts in December. We’re currently in the middle of a project to add seasonality functionality to Xmrit; if you’d like to follow along I recommend you subscribe.

Setting Up for a Sequence on Asian Conglomerates

As time consuming as the data stuff has been over the past year, the big research arc on Commoncog in 2024 has really been setting up for a series on Asian Conglomerates in 2025. This looked like two big things:

1. We published a long sequence of cases on great capital allocation, including cases on Danaher, Transdigm, HEICO, and the particularly odd case of 100-year-old-manufacturer-turned-software-conglomerate Roper Technologies.
2. We closed the year with a mini-series on Power in Business: What is Power, How to Use Power, and Power and Asian Business. This was accompanied, naturally, with a new sequence of cases demonstrating power in action.

This was an extremely slow burn, but it laid the groundwork for two things:

1. First, most Asian business empires are conglomerates. In order to examine how they work, it’s helpful to have a basis of comparison against a gold standard. The sequence of cases on great capital allocators serves this function. At the end of the capital allocation series, I wrote that I wished I had commissioned cases on Berkshire Hathaway and Constellation Software, but didn’t because I found the two companies formidable to write about. (I have, uh, been unreasonably obsessed with them both for years). We may yet broach the two conglomerates in 2025.
2. Second, in order to understand the Asian operating environment, I had to demonstrate what it is like to operate in a low-trust, low rule-of-law environment. The result was a carefully-written, very short series on power. It was also one of the most uncomfortable things that I’ve ever written; I am sensitive to the machinations of power given my operating background (Vietnam), but I have never written about it publicly. It’s no accident the entire series is behind a paywall. Western operators can live with abstractions like ‘rule-of-law’ and ‘building trust’; operators in developing countries have no such luxuries. I needed to communicate this viscerally before starting on a series on Asian tycoons, for hopefully obvious reasons.

In the coming months, we will start looking at Asian tycoons and Asian conglomerates. In true Commoncog pedagogical fashion, I’ll publish one case at a time, letting you draw your own conclusions from the patterns you observe between the cases, before coming in with my own observations. In truth, much of what we publish will be behind the paywall, since I’m going to start making some uncomfortable observations about the nature of political power and business.

There is some method to the madness, though it’ll take some time to get to the points that I want to make. I know, I know, I promised this years ago. I’m finally going to make good on finishing the Chinese Businessman series.

Side Quests: Employee Culture, the Idea Maze, Incentive Design, Brand

There were a number of interesting side quests in 2024, though not as many as in previous years:

• The biggest, most important one was probably the cluster of pieces around employee culture. For the longest time I’ve wondered how to think about employee culture, having dealt with it in all the organisations I’ve run; I could tell that some people were better at shaping culture than I was. But what is the skill of influencing culture and how does one get better at it? It wasn’t until I met Stan Slap that I got a good answer. This led to three pieces: a book summary of Slap’s 2015 book on shaping employee culture, Under The Hood, a podcast interview with the man himself, and a new sequence of cases demonstrating ideas from the book. The core contribution that Slap makes is that he thinks of a culture as an organism that exists to protect itself — which is a coherent model of what might otherwise seem like an abstract thing.
• The Idea Maze is a Useless Idea — the title speaks for itself. To be read alongside the Idea Maze concept sequence.
• A number of articles about the restaurant business, incidentally commissioned because I was still figuring out a way to scale case production and decided to start with a business model that my writers understood. This includes: A Big Part of Becoming Data Driven is Will Not Skill, Running a Fine Dining Restaurant Through a Recession, Danny Meyer’s Business Expertise, Getting Burnt by Business Expansion, and The Hospitality Solution in Your Business.
• We also started a Brand concept sequence, something that I still don’t know that much about. But the highlight there was probably the case on the Brooks Turnaround (and the subsequent discussion it sparked).
• I’m also unreasonably proud of Businesses as Ecosystem Organisms, but mostly because I could finally articulate my discomfort with the study of businesses given 15 years of ZIRP (Zero Interest Rate Policy).

There are a few other articles that we published in 2024, but I’m leaving them out for the sake of brevity.

What We’ve Learnt

I mentioned earlier that Crystal Widjaja has been a huge influence on me in the context of data. I want to tell a short story about this; I promise it’ll be worth it.

In roughly Q3 last year, I was on a call with Crystal about something data-related. At the end of the call I was walking her through the discoveries we had made about Commoncog’s causal model. I was excited — the Amazon-style WBR and Xmrit had made it possible to do lots of trial-and-error experiments, to figure out how the business actually worked. I think I was in the middle of telling her, animatedly, how we doubled our weekly newsletter sign-ups, when she interrupted me gently.

“Yes, but is this the most important part of your model?” Crystal asked. (I’m paraphrasing here; she said something much nicer than this.)

I was stunned into silence. “Oh.” I said. She asked a few more probing questions, and then I realised what she was getting at — “Oh. You’re asking if this is the highest order bit.” I stopped, leaned back in my chair to stare at the ceiling, and then groaned. “Oh, god, just because I have all these powerful, fancy data tools now doesn’t mean I should just use it on the first thing I see — I should use it on the highest order bit! Because the highest order bit is the bottleneck in my business!”

“What is my highest order bit?” I asked aloud, and then immediately, because of course I knew: “I need to figure out why people are buying Commoncog, and if their Jobs-to-be-Done aligns with the future direction of the Case Library. And then I need to test if the Case Library offering is compelling to my chosen segment. And if it’s not compelling, that means that I’ll have to find a configuration where it can be.”

(Also, the unsaid bit: if I can’t find a configuration that is compelling, I would need to go back to the drawing board. But such is the game of creating new products, and we both knew it.)

Crystal didn’t say anything during my monologue, but I could tell by the way she was smiling that I was on the right path. “Ok so I need to go execute on this. I need to start a Jobs-to-be-Done interview loop.” And then I went and did that and didn’t talk to her for a few months.

This is, broadly, the story of what we’ve learnt. We’ve learnt a lot of things around how to use XmR charts and how to run a WBR, and how to discover things about a business that you’ve been running for a few years. Deming called this gaining ‘knowledge’, where ‘knowledge’ is defined as “theories or models that let you better predict the outcomes of your business actions.”

But just because we’ve gained this new capability for knowledge doesn’t mean that we should use it on the next damn thing that pops up. Much later, Crystal told me that I’d made a common mistake, one that she’d seen so many of her students and mentees do. And I thought I had internalised the lesson of the highest order bit years ago!

I suppose this is why it is difficult to learn wisdom. You can learn something ‘in your head’ in a day. It takes much longer to learn something ‘in your bones’.

We finished the JTBD interview loop in November 2024. I … how should I put this? Do I regret that we finished so late? I think ‘regret’ is too strong a word, but I certainly wished we’d learnt this lesson earlier. Instead, I executed on a handful of other things that were all somewhat valuable, but not the highest order bit:

We Invested in Our Community

This was a Q4 initiative, and one that I’m glad we did. After a number of JTBD interviews where members expressed interest in hanging out with like-minded, self-selected folks, it seemed to me like the Commoncog membership forums was an untapped opportunity. I wrote, on the event of the forum update:

The high level reasoning for the reorg is as follows:

• The quality of Commoncog’s membership is pretty amazing. Y’all are business operators, advisors, and investors of extremely high caliber and thoughtfulness.
• Social media is only getting worse — Twitter has fallen off a cliff ever since Elon bought it, and LinkedIn is … LinkedIn.
• Let’s create a third space.

To be more precise: let’s create a third space for business nerds. The thing that I’ve learnt over the past year or so is that Commoncog’s membership is united in one thing: interest in the art of business. You could call all of us ‘business nerds’ — regardless of whether we’re investors, or operators, or advisors to businesses.

I’m really pleased with the results. I still see Commoncog as my vehicle for continued business education, and I’ve learnt a lot from the membership. I think the two highlights for me were ajzitz’s introduction of a ‘Sutton market’ as an entry barrier, and michaeljon and Luke_Stevens’s notes on branding in response to the Brooks Turnaround Case.

I also want to thank member Gilbert_Lee for the prod to model the forum after one of the many long-running value investor communities. If you’re a current member: thank you. I hope to learn more from you soon.

We Learnt to Hire Writers

I spent a large chunk of my time iterating on finding, hiring, and then training case writers. We don’t have a great process yet, but we produced a lot of cases that I’m proud of, and the process did lighten my workload for part of the year.

I also learnt a lot of things that we shouldn’t do.

We’re still iterating in 2025, with the caveat that this is still not the highest order bit. I’ll also add that I’m not going to explain what we’ve figured out when we do get to a good process; this is likely going to be a competitive advantage.

(Besides: you already know how to get there — just use process control and iterate with a WBR!)

We Learnt to Write (More Types of) Cases

It’s hard to believe that it’s only been a year from when we started the Case Library. Creating a Cognitive Flexibility Theory-backed library of cases has never been attempted, so this is very much a work in progress.

I want to talk about three things, briefly:

1. It turns out that there are more types of cases than we thought. Some cases are short vignettes, designed to demonstrate a small concept instantiation; others are long and stuffed with context. We’re learning that there should be different writing and research approaches to both.
2. One challenge that we faced was “what is the standard for truth in the cases we publish?” This might not seem very important, until you consider that we’re going to start publishing cases about corruption and smuggling and war very soon. So these things matter, even if only on an operational, legal defence level. I don’t want to get too deep into this now except to note that I’ve looked into adjacent disciplines to see if there were things I could take; academic history has too high a bar, while investigative journalism is too contentious. But I think we’ve figured something out. I’ll write an official policy soon.
3. Another decision that we faced was our ‘no authorship’ policy. You may or may not have noticed that the vast majority of cases in the Commoncog Case Library have no byline. This was intentional. Of course, writers may say that they wrote this case or that case (and we will back them), but the case itself is presented without an author. (The exceptions are ‘synthesis cases’, which contain analysis for a set of other, ‘normal’ cases.) Why do we do this? It’s partially for the same reason that The Economist has no byline — Commoncog, as an institution, stands behind the cases it publishes and takes responsibility for corrections. This means the collective is responsible for whatever we publish; it also means that this constraint guides my org and system design.

One final note. I want to thank Rhea Purohit — who I consider, in many ways, to be the Case Library’s founding writer and editor. Rhea’s dedication to finding the truth is exemplary, and she’s been an excellent thought partner when working through many of these issues. A quick story: shortly after publishing our case on Henry Singleton, we discovered a better narrative history of the man (and Teledyne, his conglomerate). I sent it to Rhea, and she said “I don’t know how I missed that! I bet the author had access to the book (Distant Force) which we didn’t. Hmm! This won’t happen again.”

When working with Rhea, it often feels like we are in pursuit of base reality. It’s made building the case library more fun.

We Built Software

2024 was the year that I started to think about building a serious software development capability into the org. Some of you may know that Commoncog’s Case Library is entirely custom-built. What you might not know is that we’ve also built in-house software to run our Weekly Business Review. (This is a reflection of the sorry state of the Modern Data Stack, which seems to be more interested in navel-gazing data problems than in delivering actual business outcomes.)

Xmrit was very much a byproduct of this capability. Will we release the WBR software we’ve built? Maybe. I’ve shown our WBR platform to a number of folk in other companies, and nearly all of them ask “how can we get a copy of this?” To which I respond: “well, how much are you willing to pay?” and then I wait to see if they’re serious about running a WBR, which is a heavier lift than just software. None of these inquiries have panned out. And so, right now: this is not the highest order bit.

There is another reason to want to build a software capability, though. As some of you may know, there’s this new-fangled technology that’s going around that generates globs of text. That technology seems rather useful if you’re in the business of slinging globs of text. But it doesn’t seem to be as powerful if you don’t build some system around it.

I don’t have much to say on this note, and will have less to say if it doesn’t pan out. But if it does, I’m not that likely to explain what we’ve learnt or built: plus, you already know how to get there — just use process control and run a WBR!

Looking Forward

What is there to look forward to in 2025? I think the bulk of the year will be taken up by the Asian Conglomerate series, which excites me. There are things in my operating environment that I’ve wanted to talk about that I can’t (for years now); we are going to go on a business adventure together.

But what else, apart from that? I am a little wary of discussing what’s coming down the pipe, especially given all that has happened over the past year that derailed my promises in the 2023 recap. But, tentatively:

• You should see big improvements to the Case Library and the core Commoncog offering, as a reflection of the insights we’ve gathered from the JTBD interview loop.
• You should see some new content on Cognitive Flexibility Theory, which is the learning science that underpins the Case Library.
• We’re going to talk about the nature of Demand, with the help of some friends.
• It’s highly likely that we’ll publish a number of cases on protocols that I teased early last year.
• And I want to spend some time to talk about Process Power, which is potentially useful given all the data-driven and Deming-related work we’ve put out recently.

Otherwise, stay tuned. We’ve got an interesting year ahead of us. I can’t wait to see what we learn together.

Originally published 08 January 2025, last updated 08 January 2025.