The Case Library Alpha Test

Here’s an email I just sent out:


Hi there,

You’re receiving this because you’re a Commonplace Member.

Bottom Line Up Front

I’m launching a Cognitive Flexibility Theory (CFT)-inspired business case library as part of Commonplace. I’d like you to help me test it.

What you’ll get: you’ll receive eight emails on one concept — one a day, over the course of eight days, starting next Wednesday (1st June).

The first concept is going to be about scaled economies. The first email will serve as a short explanation of the concept, along with a summary of CFT learning theory. The next six emails will be six different cases illustrating the various instantiations of scale economies in the real world. The final email will ask you for your feedback.

You’ll also get periodic updates about building the CFT library after the alpha test is done.

Longer Explanation

CFT is a theory of expertise in ill-structured domains. I first talked about CFT in How Note Taking Can Help You Become an Expert. To recap, there are four big ideas in the theory:

  1. An ill-structured domain is a domain where there are concepts, but the way these concepts instantiate in the real world are highly variable.
  2. Because of this variability, cases are more important than concepts when learning in ill-structured domains.
  3. Experts in ill-structured domains reason by combining fragments from previous cases that they’ve seen. In other words, they reason by analogy.
  4. Experts in ill-structured domains have an ‘adaptive worldview’, meaning that they do not think there is one root cause or one framework or one model as explanation for a particular event that they observe in their domain.

The authors of the theory then detail a method for accelerating expertise in ill-structured domains. They say that if you present a learner with a concept, and then give them 10-20 cases that are very different from each other, the learner will rapidly internalise the feel of the concept, and would be better equipped to construct fragments from the cases or recognise instantiations of that concept in the real world.

Which got me thinking: What if I did this for business? What if I built a CFT library around all the business stories I’ve collected?

In truth, I’d already noticed what any suitably voracious reader would’ve noticed: if you read a large number of business biographies, certain deep structural patterns would begin to show themselves, even in very different industries, at different time periods, in books written by very different authors.

These concept instantiations may seem ridiculously different on the surface. But they all tend to have the same ‘feel’ to them, and if we take CFT seriously, then we should trust in our brains’s ability to reason from case fragments when we are asked to make decisions in business.

Collecting those case fragments appear to be key to such expertise. The question is how to get those case fragments if you don’t have the time to read.

This is where we might be able to help.

The Commonplace Case Library

Here’s the idea: we build a case library for you. The library lives on the Commonplace website, but members are able to receive cases for a concept as an email sequence if they see fit. They may also go to the site to browse cases or search for concepts that they want to dig into.

This library will start out sparse, but as time goes by, the value of the library should compound, since sufficiently complex cases tend to be rich with multiple concepts. (This is already true based on the few cases we’ve written).

As a reader, there are two ways I think you might view the case library’s value to you:

  1. The case library will serve as a way to grow the collection of fragments in your head, without spending time on a large selection of books. So, for instance, if you want to read stories about successful org design, you’ll be able to browse a library of fragments that are drawn from say ten different business biographies. You won’t have to hunt for these stories — which in any case may be prohibitively difficult: in many cases, these books might be written by writers that did not think they were describing a valuable concept instantiation (Kochland, the story of Koch Industries, springs to mind here — there’s a particularly interesting sequence where Charles Koch reorganises the company for fast adaptation to oil price volatility; if you were looking for good stories of successful org design, you might not think to look there. Similarly, Sebastian Mallaby’s The Power Law describes venture capital org structure, and how some firms are built to last while others are not. A case library brings these patterns together for you, and saves you from having to search for each book separately).
  2. Secondly, the case library will serve as a reminder that concepts are not as clear-cut as framework descriptions make them out to be. I’m hopeful that I can link out to and embed certain cases as time passes, because every framework is flawed and oversimplified in some manner. I also hope I’ll be able to discuss cases in the members forum with you — and that you’ll be able to point out concept instantiations that would never have occurred to me.

This is the goal, and I think it’s a good one. But there remain many challenges.

The Challenges (or: Where I Try to Set Expectations)

Here are several challenges that I’m working through, as I set out to test this concept:

  1. How do you capture subtle, non-obvious concepts? (For instance, the concept ‘launching a second product is surprisingly precarious and may be dangerous for the business’ is a pattern that is difficult to describe, but shows up often enough to cause me to pay attention).
  2. What software should I use to build the case library?
  3. How do I accelerate and scale case creation?
  4. And, finally: can I build a sustainable business around this, so that the case library can grow and its value may compound?

I don’t know the answers to many of these questions, but that’s where you come in. The goal with this Alpha test is to answer the following set of questions:

  1. What is the ideal length and writing style for each case?
  2. Is sending one case a day too much for most readers, even if they’ve opted-in for it?
  3. Is it better to call out concept details explicitly? Or is it better to have readers identify embedded concepts by themselves? And finally:
  4. Would CFT’s prompts work well in an email format? (e.g. ‘How is case A different from case B, which appears to be very similar?’)

Your feedback will also help me with the ‘scale’ question — how do I hire for, train, and help writers accelerate case creation? After all, I’ll need to figure out a repeatable method for building the case library, without sacrificing quality.

Finally, I should note that I’m not doing this alone. All of the cases you’ll read during this Alpha test would have been written by Commonplace’s first intern, Fung Guan Jie (@guanjief), who will be spending his summer working on the case library project.

You’ll hear more from him as we make progress on the project.

Thanks, and see you on the other side!

Warmly,
Cedric


This is the forum thread to discuss the Alpha.

6 Likes

Case #1: Ford


Very interesting case! Here are some of my thoughts on this:

I think this demostrates scale economies in a bit of a different way than the textbook method. By the textbook method, I mean some combination of lowered costs from bargaining power with suppliers and ability to amoritze out fixed production costs over a much larger volume.

What Ford seems to be doing is finding economies in variable costs because of their sheer volume. With COGS being, of course, a multiple of Unit Cost x Number of Units Produced, Ford’s equation is dominated by the right hand side. This means, any small reduction in Unit Cost will have a huge impact on their COGS. And because it’s so large, it makes sense to invest in a huge fixed cost that will only slightly lower their variable cost.

Another interesting insight was that by reducing their price, they increase the volume, allowing them to eventually get a better cash flow as they can better power these cost savings. (The tricky part is how to know that volume will increase enough to enable this?)

This is a bit of a moat because as long as they can capture market share, they can continue to reduce costs in this way. But there’s going to be a natural limit as you can only reduce costs so much, and other companies will find their own ways to lower their costs, too. Ford’s undoing appeared to then be his stubbornness in not being nimble to buyer’s evolving tastes. Eventually, the low price of the undifferentiated Model T was not enough of a value-prop to maintain market share and their volume-based flywheel began to crumble.

This reminds me of a case with John D Rockefeller in which he would inspect refineries for cost savings. He saw somebody sealing an oil barrel for delivery. He asked how many beads of solder he applied to the barrel, and asked if would be just as effective with one fewer bead. Sure enough it was, and this one bead of solder saved a boat load of money across all of the barrels Standard Oil produced.

I haven’t quite figured out what the story of Fordlandia is trying to tell us. Perhaps Ford was so myopically focused on in-sourcing all raw material production, thinking it a magic bullet, that they tried to do something as foolish as manage a foreign jungle-bound plantation – clearly way out of their element.

1 Like

:smiley:

I think it’s worth it to remember that automobile companies were basically hypergrowth tech companies of their time — massive growth rates, since it was a new industry and everyone and their uncle wanted a horseless carriage.

Heh :grimacing: Without addressing this directly, I should note that while sometimes we include things for a reason, other times we include things for no reason, partly to see what people pick out as a pattern, and partly to capture the messiness of reality. (It’s all real, and part of the history after all!)

As it is, you’re already noticing certain things that I and @guanjief did not, so bravo!

PS: re: Fordlandia, I won’t reveal whether we included it for a reason or not — that’s left as an exercise for the alert reader. :wink:

1 Like

I haven’t read any other replies here yet; I’ll jot some thoughts down and then update with replies once I’ve shared my views.

Similar cases: IKEA, where they start with a price before designing the product; Chick Fil A, where the nuggets are made from scraps from trimming breast pieces for sandwiches; Walmart, where they exert ruthless pricing power on their suppliers (not quite the same as a vertically integrated supply chain); Apple, where they’ve focused recently on buying up small businesses toward building an integrated supply chain (especially when it comes to semiconductor design); IBM and the micro channel architecture, in the area of “building a thing to try to preempt weakness and failing at it”

Differences: so much current focus is about disaggregating integrated companies. HP split up long ago, IBM has sold many parts of itself, GE os splitting in 3, and it’s now considered routine to build directly for AWS without trying to maintain any owned infrastructure. There will of course be scale benefits for any company to run their own infrastructure directly…if they can manage it, and a company only builds the tacit knowledge about where to improve their efficiency by starting with what they have. If Ford had designed their plans around “punch outs that can be repurposed into radiator caps”, they would have missed other scaling opportunities.

This also encapsulates the innovator’s dilemma, and the core agile vs. six sigma trade off. Choosing to get really good at doing things a certain way blocked their ability to do different things cheaply. It’s easy to focus on what they could have done to build a better car, but car building wasn’t their competency, it was scaling, and they might have done better to find ways to apply their existing processes to novel fields (like bicycles or boats or home generators).

Fordlandia is interesting also because it’s an obvious failure in hindsight, and has become legendary in a “Jurassic park” sort of chaos theory sort of way - but also, it could have worked, if they’d treated it more like an incubator and less like a solved problem. (I’m writing this from an area of the US east coast where people spent centuries trying to figure out what they could grow here…before they eventually discovered how to succeed with rice, and indigo, and cotton, albeit only with forced labor). Rubber production was a big fear in World War II; one wonders if the development of synthetic rubber would have proceeded if Fordlandia had been more successful.

Finally, this touches on something I got thinking about in my newsletter this week - about the gap between between safety and monomania in creating space for innovation. Ford had a tremendous amount of runway early on due to Henry’s monomania, and this gave them space to experiment and explore other ways to improve their margins. But the same monomania locked them into a path that they had trouble leaving. In the end, it wasn’t six sigma that held them back, but their founder’s stubbornness.

1 Like

I think people don’t often recognize just how much potential the internal combustion engine released - for folks like Kohler, or the Wright Brothers, or Ford, who all realized that the IC engine solved myriad problems in applying force where it was needed. (The Wright Brothers realized that the problem of flight was control surfaces and power, and that the IC engine would let them provide power in a way that would let them study the control surfaces better.) It’s similar, I think, to the insight that people had in the 90s about networked computing or in the mid nineteenth century about steam power.

I remember a comment from the original iPhone launch to the effect of “this will still be a huge win even if we only get 1% market share”. Apple was used to low market share. But there was no natural ceiling to the number of sales that could be made.

1 Like

Harold Livesay in AMERICAN MADE (Boston: Little, Brown and Co,1979) pg 174
" Once he had designed his utilitarian, universal car, Ford “froze” the model to avoid the expense of retooling and set out to reduce the costs by increasing manufacturing efficiency, The result, of course, was the automobile assembly line, first developed by Ford and his aides, Sorenson, Knudsen, Pete Martin and others. By trial and error they developed the system by dragging the chassis across the floor to stations where parts brought by pulley, conveyor, or inclined plane were bolted on…
" … the assembly line evolved slowly throughout the car’s eighteen year production run. Ford applied four basic principles to increase efficiency: the work must be brought to the man; the work should be done waist high so as to eliminate lifting; waste motion, human or mechanical, must be minimized; each task must be reduced to the utmost simplicity. The result was a prodigy of production and cost efficiency and the obliteration of skill and the worker satisfaction that went with it.

1 Like

Going to try to add some observations not already made:

Fordlandia seems to be an interesting pattern of non-adjacent extension that tends to backfire in almost all of the cases I’ve read about. It backfires when principles that naturally evolved and worked in one place are applied to a completely different context and are transpored without any change for context. So there are a few interesting things to me here:

  • (1) When economies of scale production are first introduced, the ‘optimal state’ was reached through a period of iteration. That iteration (call it evolution?) allowed scaling up to go down the most ideal path. It wasn’t premature optimization, it was optimization of tried and true things. Meanwhile, Fordlandia was a wholesale graft of a pre-optimized system onto a cultural context and geo/bio conditions that weren’t adapted for it.

  • (2) This reminds me of Kirkpatrick Sales’ “Rebels Against the Future” about the Luddite movement in the UK – the pattern I’m noticing is that there’s an “Overton Window” of acceptable practices that you need to start at and incrementally evolve out of so as to not have workers rebel against you. Another thing this brings to mind is James C. Scott’s “Seeing Like a State”, particularly the chapter around natural vs. state-planned rational forestry. My takeaway is that the “adaptation” and “evolution” periods also allows you to optimize for local idiosyncrasies. This is true for both Fordlandia (need to ‘read the room’ and adapt to the local geo and sociopolitical contexts), and also to the scaling and insourcing. I wonder, if they had insourced too early, if they would have found the same efficiencies?

  • (3) I read Peter Chapman’s “Bananas” about the United Fruit Company which also mentioned Fordlandia. It’s been years since I read it, but I remember a large section being devoted to fruit-blights due to monocultures. This also recently happened to Florida oranges, which (from what some Floridians told me) are now mostly sourced from California. If you look at Ford’s “Any color as long as its black” and removal of customization to lower costs, then the dots all seems to draw a circle around an idea of monocultures being bad: they generate efficiencies in the short run, but are liable to existential disruption in the long run.

  • (4) I think that some scale economies – particularly those around production processes – tend to create a Sword-of-Damocles-type noose around a business: the more you optimize your specific production line, the more you’re committed to that production line and the more expensive it is to move away from it due to sunk costs. The scale effects tend to imprint their own logic on a company’s operations that makes it hard to move away from them. That in turn suggests to me a need for healthy R&D and diversification when the core business is a monoculture. To some degree, I’m edging towards the conclusions of Clay Christianson’s “Innovation Dilemma” / theory of disruption, but I’m also leaning towards the Bell/3M/Google-style R&D as a risk-mitigation strategy. (Or if you’re a fan of the evolution analogy, the value of random mutations in trying out potentially valuable diversities over time). That said, while those labs have had some really cool outcomes, I’m not sure if I know of any great studies of how R&D labs were used effectively for real business risk mitigation. So maybe I’m using some wishful thinking here.

Lastly, in terms of differences, I came across this article recently about Toyota’s “Just In Time” manufacturing: In Case of Emergency — The Prepared. I think Toyota’s an interesting counterpoint in that they are the same industry, but a different development of practices and efficiencies. In terms of the “local optimizations” thinking above, the following was an interesting local idiosyncracy which I did not know about at all, and has changed how I think about Toyota (and respectively about Ford now):

Hall considered the history of supply networks in Japan, noting how fundamentally different it was from the network that had developed in the United States. Here he makes a distinction between “type I” and “type II” suppliers. Type II suppliers, he explained, have “skills not possessed by the customer and not easily developed by them”; they are experts in their respective areas, and as such they work with industrial customers to solve problems. They are, in other words, partners in areas the customer does not specialize in. Most American suppliers, Hall suggested, were type II suppliers.
Most Japanese suppliers, on the other hand, had begun as type I suppliers: They were distinguished primarily by their capability to deliver materials at lower cost. This was either because they had lower labor costs, or used more specialized equipment, spreading its cost across multiple customers. In Japan these suppliers had been inextricably tied to the development of large firms like Toyota, and those firms often directly loaned money for their creation while maintaining exclusive or preferential arrangements. These were subcontractors, in other words. This was not a network built on partnerships between equals, but with subordinates in a supplier family.

Definitely thinking about that re: the insourcing that Ford was doing.

The related string of thought here is that for anyone who has worked in a management or leadership role in any business, you know there’s an inherent trade-off between insourcing versus outsourcing. This is as true for tools and materials (vendors vs. in house) as it is for people (outsourcing/agencies vs hiring staff). While you’re in a time of uncertainty, it makes more sense to outsource even if the costs are higher in the short term because it creates optionality and flexibility. Its easier to change vendors and their skill sets than it is to train someone internally to do a different job or to rebuild a casting process.

The Prepared also goes on to describe what mass-production is and isn’t:

Mass production is sometimes called continuous production, and by every measure this is a more accurate label. It is a way for assembling one part after another, performing one process after another, with each stage of production proceeding in sequential operation. It is about flow, and the control of flow. Indeed, what struck observers of the Toyota system was not that it never stopped moving—but that it stopped moving all the time. “Jidoka” meant that the assembly lines were “equipped with stop buttons at each station,” and workers on the line were told to push the button “anytime they were unable to finish a task.” Usually this indicated that the line’s operation might need to be revised. But it also demonstrated a responsiveness to the individual elements of the system. While the economics of setting up an assembly line may favor a certain scale of production, its flow has always been the primary variable of interest.
[bold emphasis mine]

This is an interesting way to build responsiveness into mass production… I wonder if it extends to not just production-process responsiveness, but market responsiveness? (And I think the answer to market-responsiveness was Just-in-Time manufacturing).

As a meta note, should this case study have its own thread for discussion?

1 Like

This reminds me of a more recent example of how large supermarkets chains cut costs by simply shortening their receipts. Again, making such a small change means a tiny cost saving per receipt, but since they print so many receipts annually, Walmart said it saved them US$7 million in 2017. A smaller supermarket chain, LIDL UK, did something similar and saved £150,000 annually too.

I think that was the view Greg Grandin, the author of the book Fordlandia, held too. He said in a NYT article:

Ford was obsessed, among other things, by Thomas Edison, soybeans, antiques and order. He hated unions, cows, Wall Street, Franklin Roosevelt and Jews. He also, fatally, despised experts. Ford’s Amazon team had plenty of able men, but as Grandin observes, “what it didn’t have was a horticulturalist, agronomist, botanist, microbiologist, entomologist or any other person who might know something about jungle rubber and its enemies” — the lace bugs and leaf blight that laid siege to the rubber trees, the swarms of caterpillars that left areas of the plantation “as bare as bean poles.” (emphasis added)

Yup! Ford was fixed in their ways of mass-producing the same model, the Model T, while General Motors had a different policy to produce a range of cars, one for each price range. I found this fascinating passage by GM CEO Alfred P. Sloan’s book, My Years with General Motors:

In 1921 Ford had about 60 per cent of the total car and truck market in units, and Chevrolet had about 4 per cent. With Ford in almost complete possession of the low-price field, it would have been suicidal to compete with him head on. No conceivable amount of capital short of the United States Treasury could have sustained the losses required to take volume away from him at his own game. The strategy we devised was to take a bite from the top of his position, conceived as a price class, and in this way build up Chevrolet volume on a profitable basis. In later years, as the consumer upgraded his preference, the new General Motors policy was to become critically attuned to the course of American history. (emphasis added)

I’m thinking that for this Alpha, we should just do everything in one thread, to make it easier to call out to previous cases and discussions on previous cases. This may or may not be a good idea — but let’s try this as an experiment for the duration of the test.

:upside_down_face:

So this observation is amazing. This is so, so true — in the first few chapters of My Life and Work, Ford details how he created the Ford Motor Company, along with all the precursor companies he started that failed along the way. It’s pretty remarkable that he didn’t take the same iterative approach he took to his business and his production lines to the rubber plantation project. The man is a complex one, for sure — My Life and Work is 100% reasonable and well-written and thrilling because you see him working out the principles of lean production from scratch, and it’s all well and good until you reach the anti-semitic final chapter and go ‘Wait, what?!’

This bit was great — you’re right, I hadn’t thought to compare the ICE to networked computing in the 90s.

One thing that we didn’t manage to put into the case study was the fact that Ford had huge amounts of employee churn thanks to the drudgery of the production line, until he decided to raise wages well above what all his competitors were paying, and reduced the number of hours worked, at which point absenteeism and churn suddenly vanished.

I guess the Ford of today would be harping on and on about the 4 day work week :wink:


I’m really enjoying the discussion so far — this just highlights for me idea 4 in CFT, that:

Experts in ill-structured domains have an ‘adaptive worldview’, meaning that they do not think there is one root cause or one framework or one model as explanation for a particular event that they observe in their domain.

I mean, it’s just very clear that there are way more concepts embedded in this case than @guanjief and I thought there were, and perhaps this is an argument for making cases richer and messier than is strictly necessary.

In the first set of questions before Case #1, my thinking on scaling is that scale only matters if what you have scaled continues to matter to and remains relevant to your customers. If you don’t adapt or find new opportunities, customer sentiments will change, market dynamics will shift, and your advantage will erode.

SpaceX is a current example of a company attempting to build economies of scale in an industry that hasn’t achieved scale yet. They have focused on reducing costs by inner-sourcing certain rocket components and through rapid experimentation established reusable rocket components. These optimizations have resulted in more frequent launches resulting in more cash flow, etc.

When I think of Apple, I also think of what happened to Blackberry. Just 12 years ago Blackberry had 43% market share of mobile phones! Only 3 years later their share had dropped to 5.9%. And in January of this year, they shut down all Blackberry services for good. Like Ford’s stubbornness in only offering black painted cars, Blackberry was stubborn to move away from a physical keyboard and it cost them dearly. But unlike Ford, Blackberry wasn’t facing similar competitors who were catching up with cost savings and industrial production line operations. While I never owned a Blackberry, I can almost guarantee it was easier to type with than a touchscreen phone is today. But the world around Blackberry was evolving to a point where typing wasn’t all that you wanted to do with a phone, and they either didn’t see it coming or didn’t anticipate the speed at which market dynamics were shifting against them. They were displaced by Apple, a computer company, and Google, a web search advertising company.

Regarding Fordlandia, 99% Invisible has a podcast about it here if you want to learn more.

2 Likes

Neat start to the case studies!

His use of thresholds at the beginning of the case sticks out most.

Publishing ebooks worked similarly. If our CPC was $1 and our earnings per click (EPC) were $0.90, the overall profit was going to be negative and we’d abandon it. However, if the CPC was $1 and the EPC was $1.10, we knew we could scale the offer to millions of dollars in revenue.

1 Like

Case #2: Netflix


Your thoughts on the Netflix case goes below. :wink:

Similar to: there seems to be an overlap with “starting with investing in scale” rather than benefiting from scale where it appears in a way that mirrors Ford, though with Ford the goal was hitting a cheaper price, not a higher subscriber count. Trying to add more value at an existing subscriber price point is a lot more like Disney world, a lot less like manufacturing.

Differences: they were not just iterating on an existing product like the model T; it was more like transitioning from a reasonably successful initial product to something with more potential longevity. They could have instead doubled down on scaling and vertically integrating their DVD business (for example, buying Red Box, getting cheaper residual broadcast rights like Comcast did with On Demand)

They pursued a much, much less aggressive vertical integration than Ford. They started buying content directly, and took a more active role in commissioning it, but I don’t believe they ever stepped in to making it directly themselves. Their vertical integration at the end was a lot more like where Ford was when it started. Perhaps their suppliers were more clearly Type II, while Ford’s were type I, to use @roman’s framing above.

I keep thinking about this recent stratechery post about how Cable came out ahead post cord cutting. As much as Netflix’s activities were good for subscribers, they were awesome for cable companies and movie studios, and it’s a bummer for Netflix that they didn’t see more residual from that.

Also not mentioned in the case, but relevant: Netflix attributed a lot of their virtue in original content to their deep customer analytics and machine learning systems. I don’t know whether it’s actually the case - but they publicly framed this strength as something available to them only due to their scale.

1 Like

By the way - I’m on vacation this week, and this has been a welcome intellectual diversion while I sit on the beach. We don’t usually eat out much, but have been getting takeout each evening, and I daily marvel at the disadvantages businesses here are facing due to their inability to scale. (Most of the nearby seafood restaurants have lines that take 1-2 hours to get through. Lots of customers come and go during that time. They would do substantially better if only they had a few additional kitchen staff, even without extra kitchen space, and the labor costs would probably cover themselves.

Anyway…as I wait in line every day, I think about scale.

3 Likes

There are a few interesting things that stand out to me:

The line about amortizing cost of production across a base of viewers as a differentiator business model (often compared against Hulu) made me consider what the other options are:

  • (1) Apple TV dropped considerable investment into becoming an HBO of streaming as well (what feels like to make Apple TV the preferred hardware choice, but now they’re mostly just another channel… not sure what the long term strategy is here)… and used that huge investment to build a user base… aka production as customer acquisition.

  • (2) Disney already had high-value IP and so their customer acquisition was low based on an established brand, but they had hard limits to the volume of content they could pump out without ruining their brand and oversaturating their IP… so they used their considerable heft to purchase new IP and squeeze extra production out of it (Star Wars, Marvel, ESPN, Discovery), and then expand that to other branded content. (Places, things).

  • (3) Amazon followed Netflix’s licensing model first, but gave away their video content free as part of Prime. It was a value play for Prime, but also uses amazon video as a portal to sell more video (I think only Apple TV also sells video on demand of the big players). So it’s both a streamer and a sales channel in itself.

Like @peter , I’ve been thinking about that recent Stratechery post, but I’ll try to take it a step further. There are a lot of interesting scale-related things, but the thing that really jumps out to me is that video production (shows and movies) is in itself something that requires a certain kind of scale. Look at the credits of almost any modern movie. Even the more indie productions nowadays (horror movies excluded) have a budget of 10-20million at the low end and require hundreds if not thousands of people.

So Netflix investing in movie production is a huge investment. It’s cheaper to build out the software and technology than it is to build out marquee movie/show production capabilities. But once the coat of technology is low enough, IP holders can have their own channels (isn’t that what these are nowadays?). Hence Peacock, Showtime, even the doomed CNN+ pulling their IP and trying to launch their own channel. Way fewer companies are moving in the other direction - AppleTV, Netflix, and Hulu are the only ones that stand out - we understand Netflix, Hulu was absorbed into a mostly flexible business catch all for Disney, and AppleTV makes no sense to me.

Anyway, thinking about the scale of movie production against something like the production and distribution of books and music (video games will be a different can of worms). For Music, this Substack (https://tedgioia.substack.com/p/record-labels-dig-their-own-grave) kind of lodged itself on my head:

  1. Record labels have lost their ability to launch new careers. 2. Like Bartleby the scrivener, they really prefer not to deal with this whole issue because career development is such a hassle . 3. So they demand that musicians build their own audience via TikTok and other social media platforms. 4. But the moment musicians become capable of doing this, they don’t need record labels anymore.

Meanwhile on the Farnham Street Knowledge Project podcast, author Hugh Howey (Hugh Howey: Winning at the Self-publishing Game) makes a similar case for how publishing and book distribution is changing and why self publishing and avoiding the distributors makes better financial sense because what do distributors actually do for you nowadays?

Which brings me back to Stratechery (different post, can’t find it right now) where Ben Thompson talks about it really being a unique content and distribution game - that these are just the same old content games as before, but it took forced innovation to come full circle.

So then, what is the interesting thing that Netflix did? I think it was on being ahead of the curve, multiple times, and leveraging the gains from one cycle to feed the next cycle.

(1) They built a huge audience, direct relationship with that audience, and consumption profile with the DVDs business at a time when Blockbuster and Best Buy (or movie theaters) were the only options.

(2) They leveraged that audience ownership to launch a novel new business model of streaming, way ahead of the curve. (I remember reading an interview with Reed Hastings where he described their constant fear that Blockbuster would do it first or jump in… and they never did).

(3) Then they leveraged that heretofore unproven streaming business model to license IP at low prices because studios didn’t think that would ever be a meaningful business.

(4) They then started producing their own media, knowing that others would catch up and were able to amortize the costs of production against a large paying base, while also creating differentiable content that others didn’t have, and also leveraging their knowledge of real viewing habits to make each production investment work better, because they had a much tighter feedback loop on what content worked and what didn’t work… something that I think gets lost in the talk just about costs amortization. (They also gave considerable freedom to creators to try things out in the beginning, which itself drew a lot of unique creative production into Netflix rather than to other channels.

(5) … which brings us to where we are today, where Netflix is one fish amongst many, and we are seeing them struggle (subscriber loss in recent earnings call) against a lot of old timers that caught up.

I’m not quite sure the scale-related lesson to learn from this, but there’s definitely a lot here that I haven’t had time to really think about.

The one thing that really don’t understand though, is why debt financing at such large sums? Debt financing tends to be more expensive. Was this after they went public? Was this because it had no strings attached, business-model wise?

1 Like

I suppose it might not be as clear from the case alone, but we actually decided to look into Netflix after Hamilton Helmer referenced it in 7 Powers (link goes to the relevant section; it was the marquee example in his chapter on Scale Economies).

I realise now that it’s not so clear how the Netflix case is related to scale. :frowning: I’m so sorry. We’ll have to rewrite. cc @guanjief

The point Helmer made in his book was mostly that Netflix took an expensive variable cost (licensing fees to content producers) and turned it into a fixed cost, which could then be amortised across their entire subscriber base. Otherwise, they’d have to continue paying licensing fees to content producers, giving the producers leverage.


I’m not entirely sure if it’s clear what’s behind Netflix’s recent subscriber loss. Certainly, there’s a narrative that has emerged today about how ‘oh their content sucks’ — but is this really true, or a case of narrative chasing the price?

One thing that I’ve drawn from Netflix’s more recent adventures, though, is just how difficult it is to ‘turn the ship’ when you’ve got massive fixed cost spending justified by your scale — though perhaps for a different reason from your ‘Sword-of-Damocles’ point about Ford’s sunk costs earlier in this thread.


Paypal

Discussion for the Paypal case goes below. :wink:

This case seems a lot more narrowly focused than the others, and the economies of scale strike me as more of an emergent afterthought than something they planned for - especially when it comes to increasing costs for their competition.

The case reminds me of the situation Mailchimp found itself in when they launched their freemium product about 13 years ago. They started getting clobbered with fraud, but they kept up with it and got good enough at detecting it that they were able to command the low end email market for the better part of a decade. And this had some interesting downstream benefits, including being able to do low cost experimentation on low value free customers.

1 Like

I think AppleTV is Apple’s android. It’s less of an offensive move into the streaming space, and more of low-variable-cost barrier to being displaced by other competitors…perhaps similar to what they’re doing with Safari or Apple Maps. Not quite to the extent as Ford with bringing everything in house, but retaining the option to do so if they ever need to.

1 Like

I really enjoyed the PayPal case study. I think the benefit of scale here is the exact opposite of Ford — Ford built a rigid production line that was really good until the moment it wasn’t, and then it was too inflexible to adapt (though Ford did eventually adapt and they’re still around as a very successful company). One thing I started to question when thinking about that case was, “how do you build resilience into your scale model?” (I thought Toyota’s Just-in-Time manufacturing as a process was an interesting solution, but it felt incomplete… 3d printing as a technology feels like an interesting solution from the other angle, but its still early there).

Meanwhile, PayPal was forced to build resilience into their scale economy. @peter, you write that it seems like “an emergent afterthought […] especially when it comes to increasing costs for their competition”, but when I was reading it, I framed it really differently. As the Levchin said, “PayPal [the consumer interface] was more or less a commodity business” (brackets mine). The front-end commodity business was a Trojan horse, IMO. On one hand, banks — by nature conservative — didn’t really want to jump into that early and unproven space, so they left the consumer side open for someone to jump in, and PayPal was it. On the other hand, it was a hard problem to solve which is why there was opportunity here.

The commodity consumer interface was a Trojan horse/forcing function because the defensible business moat was the hidden complexity beneath the surface, which didn’t stay static — the moat/complexity continued to grow as PayPal addressed it. [1] By having a large consumer base, they increased their attack surface and were forced to confront more and more risks. And because they had lead time with no one else in the market, their head start kept getting bigger and bigger — hence Cedric and Guan’s “Poisoning the Competition”/increasing costs of entry to the competition. Not only that, but that also builds up a trustworthy brand and it allows them to keep pricing low for consumers (revenue comes from having less “leak” via fraud, rather than extracting more from consumers).

So the lesson I am taking away is that in high-risk and dynamic market sectors, scale and resilience/risk management go hand in hand. @peter your note on MailChimp is a great example of that as well. [2]

So if I were to reflect on the three case studies so far, I’d say:

  • Ford is an example of an initially low complexity, low dynamic environment with commodity products where scale could be built up in a fixed fashion — doing more of the same, but using scale/growth to drive unit costs down. This model fails when new consumer trends mean that your upfront investment actually prevents you from adapting due to sunk costs. By the time you realize you need to change and implement that change, the competition has already eaten significant market share away from you. I think of Tesla as a modern day analogue — from horses to the model T, from internal combustion to electric motors.

  • PayPal is an example of high complexity, high dynamic environment, where scale builds up resilience. Both PayPal and Ford were able to build temporary moats in terms of high-cost to enter (building a factory to produce new cars is a huge investment, limits number of people who can enter a market in a way that, say, production of cereal or shoes doesn’t have the same sunk investment costs), but Ford’s was rigid while PayPal was flexible. This grew PayPal’s moat. Because the complexity is the the backend and the consumer interface is a commodity, I think PayPal could have pivoted to off be doing what Stripe and remittance companies are doing nowadays as well. I actually think that Stripe is a similarly great case study here as well — it’s what PayPal potentially could have been if it wasn’t bought by eBay. So I think that this is where PayPal’s scale model went “wrong” — they made their backend robust, but let the consumer commodity portion stagnate, which allowed new competitors to enter and eat what should have easily been their market share.

So what then of Netflix?

My point wasn’t in trying to say that content sucks — just the opposite, I think that content differentiation is the only thing [4] they could have done to remain a defensible business (and they’ve gotten smarter and better and content production due to having tight feedback loops between production and consumption). My questioning was around that if you poke at it, there were what felt like more than a few scale-related lessons beyond the financial one, as well as operational curiosities. Being able to turn variable costs into fixed costs as a function of scale was definitely interesting and unique — I guess my point was more so that even though they managed to do that, they just recreated the studio production model. Meanwhile, studios realized what Netflix was doing, and recreated the streaming model Netflix has pioneered.

The end result is that Netflix and major studios converged on the same strategic location: you need to own great IP and you can improve the value-generation of your IP by having a tighter feedback loop to consumption. That allows you to amortize costs across an established and committed market segment and also to diffuse risk from by knowing consumer preferences better — what people actually do, vs what they say they do, or previously relying on critics. To paraphrase Ben Thompson for a moment, ‘At this point, Netflix’s chief problems are traditional media problems, not technology or financing problems.’ [4]

So while I agree that Scale allows you to do interesting things that others can’t, I was thinking that a competitor like Disney also used their scale in an interesting way. Their scale was a huge backlog of IP and Brand reputation, which they were able to use to launch a wildly successful streaming program that grew tens of millions of users almost overnight. Netflix used scale of audience to decrease costs of production, Disney used scale of existing production/backlog to generate an audience with low explicit acquisition costs. Or to say it slightly differently: Disney was able to also use their scale to move away from variable costs (revenue from when their stuff was shown on a channel or purchased somewhere) to fixed costs vis a vis a streaming subscription, a similar amortization Netflix pulled off.

[Edit 1: Still trying to figure out the best way to simplify the idea. I think Disney is an interesting counterpoint because they already had the scale of IP and suddenly made the IP much more valuable by pivoting to recurring revenue from Disney +. Scale of existing investment into IP, and scale of that IP and brand recognition to make their user acquisition cost near zero. Of the other streamers, only HBO was able to pull off something similar. ]

The other thing that’s interesting to me to unpack further about Netflix is in what they produce. Compared to Ford’s Model T, marquee content tends to have a power-curve of value. The most impactful content will continue to generate value for long periods of time, for both new customers and repeat customers. So when done right, marque content has a one-time upfront cost and a long tail of return over time (see: Disney). This is somewhat unlike the Ford model, where the upfront investment locks you in — what worked for the Model T production wasn’t generalized into flexibility to produce many other things.

If Ford sunk a significant portion of earnings into a fixed production line and hit gold, there is a one-time value generated from selling that product. If the product tanked, they’d be sold for scrap metal. If Netflix sunk significant portions into production and hit gold, that IP continues to pay dividends over a long period of time and builds up a moat. If it doesn’t, that IP still has value and the production process they’ve invested in is a one-time cost… they can just produce something different using the same production studio investment… said inversely, Netflix’s production process produces many different products while Ford produced only one.

That’s why for Netflix I keep going back to being able to own the audience and understand their real consumption patterns to amortize both costs of production and risks/certainty of production value.

(Hmm — as I think about it, one could describe the scales relationship between production process and output as One:One for Ford, One:Many for Netflix, and Many:One for PayPal. Is there a many to many case?)

Edit 2: What could Ford have done? Rather than scale as overfitting for a single product, they could have used scale to build out modularization. Which is why car companies today do: most car models across a single company share a base platform and design system and have interchangeable pieces. This allows for flexibility across trim levels and also across models. However, this wouldn’t have been obvious in Model T times because Ford’s challenge was marketing-making: by offering only one model and using scale to decrease costs, he made it easy to buy a car and built out the market.l away from horse carriages. Once people understood the value of cars, then flexibility to cater to customer desires became a thing.

——

[1] - I think of this similar to Uber in some ways, the way people ask “why do you need thousands of engineers to build what is essentially an app with 5 screens?” — it’s the complexity and edge cases that make things difficult.

[2] - PayPal’s execution of risk management actually makes me want to compare them to insurance companies vis-a-vis’s Warren Buffett explanation of how they have low upfront differentiation but leverage risk management to drive down costs and how re-insurers manage existential risks. Buffet also mentioned insurance is highly dependent on sales people and marketing for customer acquisition though, so it’s not an ideal comparison.

[3] - If anyone is interested in diving really deep into the complexities of payments and fraud infrastructure, Stripe’s Patrick McKenzie runs a fantastic deep-dive blog about those topics: Bits about Money (Page 1) … the payments infrastructure of Japan was eye-opening (for me as an American), as an example.

[4] - I worked in healthcare technology for nearly a decade and one thing that was really interesting to me was Oscar Health Insurance. For those who aren’t familiar, they launched a consumer-friendly insurance plan way ahead of the curve, and tried to grow competitive differentiation through a consumer-friendly interface, subsidizing wearable technology to learn more about human behaviors to lower costs, and deeply invested in the tech stack (including early virtual care experiments) to make ‘healthcare a better experience’. Where they didn’t manage to compete strongly enough was in provider networks (which traditional plans have a strong hold on) and getting into Employer healthcare options (which is where the majority of healthcare comes from in the US… or said inversely, they were only able to reach otherwise-uninsured people). They’re still a player in some markets and unambiguously proved their point about technology/data being able to drive costs down and care results up, but because of barriers to entry and how health insurance operates in the US, they had a hard time growing. More interestingly, they’ve started to use their success to “Oscar-fy” traditional insurers. I used to joke that my employer should have been positioned as “We help you Oscar-fy your health plan”, until Oscar actually started doing that. Whelp! Anyway, the point of this story — other than being an interesting case study on scale barriers to look into — is to look at Netflix and think, “ok — we will all converge on the same consumer solution vis-a-vis streaming… can I sell my tech to other providers and leverage learnings therein to grow more competitive?” (As opposed to pivoting to ads and degrading the consumer experience, as they seem to be moving towards now.)

1 Like

I want to quote from Kris Abdelmessih, who directly inspired the PayPal case:

Competitive equilibrium will mean that the casinos who can bid the highest for the “customer” is the house that can:

a) source the most uncorrelated offsets to the wager

and

b) has the biggest bankroll

In the trading business, condition A is satisfied by the market makers with the best data/analytics and “see the most flow”. A firm entrenched in both equity markets and futures markets with licenses from both the SEC and CFTC is going to be more efficient at laying off the risks it acquires from serving tourists regardless of the venue they choose to play in.

A and B will create a virtuous loop. The best players will build larger bankrolls which allow them to outbid competitors further which earns them first look at the flow which improves their models and so forth.

From: Why You Don't Get Paid For Diversifiable Risks - Party at the Moontower

In certain markets, scale players benefit from increased flow and throughput, and slowly poison the market for all the other non-scale players.

1 Like