The Ball-Bearing Theory of Corporate Decay

The Ball-Bearing Theory of Corporate Decay

Joydeep Sen Sarma
Joydeep Sen Sarma
←  Back

The Ball-Bearing Theory of Corporate Decay

Joydeep Sen Sarma
Table of Contents
Sign up for our Newsletter
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

Google has been in the news of late. Its efforts to roll out a competing product to chatGPT have been hamstrung by internal issues (despite having created a stellar new AI model named Gemini). Its struggles are not uncommon as firms age. Decay leading to mortality is common. But is it inevitable? Are all companies consigned to this fate?

IMO - No. One Big-Tech firm today is immune to this kind of decay: Amazon

This blog details, what I call, "The Ball-Bearing Theory of Corporate Decay". Some of the thoughts in this article were first expressed on this Twitter thread (and would encourage readers to leave behind their thoughts there).

The Startup

To start - it's useful to visualize firms as collections of individuals & functions. When a firm starts off & in its early phases - it's a small collection - and almost all the constituents are externally facing. The below image shows the conceptualization of such a startup as a small collection of ball bearings. Note that there are only a few balls hidden beneath the surface layers inside in this image - most of the balls are on the surface.

A small collection of ball bearings representing a startup
A small sphere of ball bearings

The survival of this sphere (startup), in a free market, is determined by the competitiveness of the balls on its surface. Hence startups must hire the very best. And most die anyway (the survivors we see tend to have had some superb people at inception).

Growth Stage

As the company grows, this conceptual sphere becomes bigger and bigger. The image below represents a considerably larger firm - still formed as a sphere full of ball bearings.

Now - there are a lot of balls hidden below the surface of this sphere. Notably - these spheres never face the heat of external competition. Typically these are functions like HR/IT/Finance. Very frequently - particularly in technology firms - these can be internal software, their development teams and the processes built around them.

A large collection of ball bearings representing a larger company
A large sphere of ball bearings

When firms are in the high-growth phase - they tend to attract very good people. And these internal functions & individuals are also of very high caliber and cutting edge. However, as time elapses - inevitably these internal functions start decaying. Because they are typically monopolies, facing no competition, with a weak feedback loop. The quality of people gradually reverts to mean as well.

Once upon a Yahoo!

A good example is a (once) famous company called Yahoo. For those who remember, Yahoo was the OG iconic Internet company. Far bigger than Google at the time (and nearly bought both Google and Facebook). Much of the software built initially at Yahoo to help it scale to hundreds of millions of users was bleeding edge. (One of the software developers who built many of the popular internal software components was Brian Acton - who later went on to found WhatsApp).

One such software I used during my brief stint there was a data warehousing software called Myna. A loosely coupled MPP database that stored extremely large datasets in flat files on NetApp filers. Needless to say the system was far ahead of its times (few companies had data of the size Yahoo did in the early 2000s). To some extent, Myna was an inspiration for Apache Hive - which I had hand in building later at Facebook along with a stellar team.

Unfortunately, that system did not keep up with the times and gradually withered and eventually died. First, it was replaced by Hadoop inside Yahoo itself (though from a different part of the company) - and later by Cloud Storage and many different Warehousing software frameworks (both Open Source like Spark & Presto and proprietary like Snowflake).

Google again

This is a good point to bring up Google again. This tweet thread tells the story. Not just about politics - but the reality of decaying internal functions (primarily HR) that seem to have eaten the company from the inside.

Competition is the Antidote

As a close observer of this phenomenon at Yahoo and multiple other firms - I ended up concluding that the biggest weakness such internal functions and software have is that they are monopolistic.

  • Internal Software and Service provider teams have unusual power to dictate terms to their host organizations.
  • External vendors are held to a much higher bar of service (whether uptime, or performance, or cost or scaling) - than their internal counterparts.
  • It is always much much easier firing an external vendor than taking harsh steps against internal teams.
  • It is also much harder to measure the value of such internal teams because of the lack of direct revenue linkage. By comparison, the prices of vendors in a competitive market are easily benchmarked against competitors.

Over time the effects of these monopolies, compounded by talent reverting to mean, start showing up. And overall competitiveness of the firm in free market starts getting handicapped. Even with lots of good people (who inevitably man the surface of the sphere)


How do firms fight this? While not articulated commonly in this manner - firms have long understood the concept of "Core-Competence". The classic way to fight this is "Outsourcing". Take your internal functions and software - purchase it from free market. (Slight irony that "core" in this parlance refers to the functions that face the external market).

A ring of ball bearings with no hidden layers

The firm's own functions and people are now almost exclusively externally focused. And there are few internal functions to decay per se. If one vendor is becoming dated and expensive, get a fresh one. Stuck with an expensive on-premise Oracle database? Move to a modern Snowflake instead. (This, btw, is why "standards" (like SQL) are so useful and vital).

So why do most of BigTech companies not follow this paradigm? There are at least two reasons:

  1. The stupendous growth of the internet outstripped the capability of vendors in free-market to supply sw/hw that tech giants needed. So they had to build their own specialized internal systems.
  2. A high concentration of talented people often also leads to a "Not-Invented-Here" mentality. Tech companies often want to build more things themselves, than they should.

I have witnessed both of these reasons first-hand - the second ones at the startups I founded.  The problem was particularly severe during initial many years @Qubole - where many tech teams wanted to manage their own big data stack. My current firm, ClearFeed, can do an excellent job of providing automation around Slack-channels - managing Ticketing, customer chat rooms, Escalations and Accessing Company Knowledge. The issue is less prominent here as we sell to non technical teams, but surfaces with more tech savvy engineering ones. (The proliferation of SaaS is complicating this since many firms are reluctant to use externally hosted software for security reasons).


Another common and completely different way companies evergreen themselves is by using acquisitions. Companies like Cisco have mastered this art. I would dare to say that in some of these cases - there's an admission that older products are almost guaranteed to decay, that internal innovation is hard - and that a much easier path is to keep acquiring younger firms that have stood the test of the free market. ie. - don't bother fighting internal decay, just inject some fresh blood periodically.

The Amazon Way

Jeff Bezos built Amazon differently. While Amazon often built their own systems - he insisted they be run as independent services. A lot has been written on this - so I will stress the less-stated:

Amazon also exposed its internal services as commercial products.

As the most famous example - the compute and storage farms used to support Amazon's e-commerce business, became an inspiration to build EC2 and S3 and AWS. And they didn't stop there. A couple of somewhat recent examples:

  • When AWS made it's own ARM chips - they didn't keep them as a proprietary HW for their own cloud services. Instead - anyone could rent AWS Graviton servers (the 'g' type instances) in the cloud.
  • AWS also exposed Graviton to full-on competition from Intel/AMD - instead of protecting it as a monopolistic provider behind some dominant service (like Ec2 or Lambda).
  • Similarly - AWS did not make high performance(density) instances only for RedShift (although many thought they would). Instead - such instance types were made available to the whole world when developed.

It's a consistent philosophy applied repeatedly.

As a result - Amazon looks, instead of a sphere, a lot more like a federation of independent services and products, all integrated - but all also competing in the free market.

A collection of steel balls floating together in the air

What is remarkable is that not only is Amazon decay-resistant - but that they have turned the idea of "Core-Competence" on its head. Anything can be a "core competence" - as long as it is exposed as a commercial product and can compete and survive in the free market. Amazon's strategy allows it to keep expanding its products and addressable markets - while not paying for acquisitions. It's ability to generate new businesses organically is reflected in it's stock price.

Zack Kanter explained this beautifully in this article on Amazon years back, particularly focused around its e-commerce and logistics components:

Open Source

For internal software - adopting open-source can be an anti-dote to monopolistic rot. When looking back at the history of some of Yahoo's early software - this may have helped them:

  • a popular OSS project may thrive beyond and outside it's creator organization
  • it is exposed to competition from other OSS projects
  • it's success in the OSS marketplace (relative to other similar software) is also a useful signal of competitiveness (that is not available inside the company)

Of course, later in the internet era - it became a formula for committers of popular open-source projects to start their own companies. Companies like LinkedIn, for example, benefited from the later commercialization of open-source projects they started (in particular Apache Kafka). There are also examples outside of Big-Tech (example Cloud Custodian from Capital One).

However - Open-Source, in this context, is a means to an end. One could still be stuck with internal software, made public, but not really benefiting from the dynamics of OSS.

Analogies from Economics

The visualization of competing entities as clusters of ball-bearings also helps understand the world beyond just firms. Countries are also similar:

  • Export-oriented industries of a country are competitive in the free market
  • But it's internally facing industries - particularly monopolies - may not be competitive

In Economics - these are often categorized as Tradable and Non-Tradable Sectors. And it is known that one of the benefits of export-oriented units (which are usually competitive in the open-market) is to induce competitiveness in their supply chain (thereby making other internal industries more competitive as well).

In Conclusion

I have been ruminating on this topic for many years now. It long predates Google's ongoing issues. In fact, I used to think Google was the exception to this pattern!

Like with any simplified model of a highly complex (multi-variate) system - this theory will often be wrong and almost always be too simplistic. Businesses struggle because of a myriad reasons - internal struggles being only one of them. Rapid technological change and business-model changes, for example, are even more common factors why businesses struggle to live forever. Firms are also often insanely competitive simply because of the quality of their leadership (and not because of structural reasons described here). Facebook is a good example. For readers: a simplified model like this is only useful when all other factors are controlled for.

Related Blogs

See all Blog Posts
TOC heading
Text LinkText Link Active
Get a Free consultation with a Support Expert
Learn how fast growing companies like Teleport, Chronosphere and Acryl Data have scaled Support processes with ClearFeed
Thank you for contacting us. Our team will reach out to you shortly.
Oops! Something went wrong while submitting the form.