The Enshittification Conundrum

By Hani Mebar | Äctvli Responsible Consulting

Remember when things used to be built to last, and you used to own what you paid for? Now everything is a verison of ‘renting’; the software is rented, the music is rented, the films are rented, your own photographs sit in somebody else’s storage for a monthly fee, and the cheap gadget that arrived from the other side of the world broke on the second use.

This is in laymen’s terms what enshittification means, and it’s followed by “the explanation is a simple single word: greed!”.

I share most of the complaint, but disagree with the explanation. Why? Because companies, in order to exist for longer than a couple years, need to be profitable. In fact, they have wanted profit in the 1960 and in the 1990 as much as they do today, so “greed” cannot be the thing that changed. Profit is how a business pays its people and keeps its doors open, and a company that builds something so good that nobody needs to come back for years, at least today, would eventually shutter its doors. This article tries to answer two questions: Why/or how did things evolve this way? And what should a company do if it wants to survive without treating its customers as captives?

Three Problems

The term ‘enshittification’ has an origin story. It comes from the writer Cory Doctorow, who introduced it in January 2023 with a precise meaning: “Here is how platforms die: first, they are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the value for themselves.”

That describes two-sided platforms, one for sellers and one for buyer (a market, really). But in everyday use the word has been stretching to cover at least three separate things:

Products designed to wear out, or to resist repair. The old planned obsolescence argument.
Ownership replaced by licences and subscriptions. Software, music, films, storage, and increasingly features inside physical products.
Goods that were never meant to last. The disposable end of online retail (think Temu and Wish).

They each have different causes and different remedies.

Planned obsolescence

The idea that manufacturers design products to fail was a conspiracy theory among die hard fans of certain products. You’d be surprised to learn however that some of it is documented history.

On 23 December 1924, representatives of the world’s largest lightbulb makers, including Osram, Philips, Compagnie des Lampes and General Electric, met in Geneva and founded what became known as the Phoebus cartel. Among other things, the cartel set a standard lifespan of 1,000 hours for household bulbs. Members submitted their bulbs to a central laboratory for testing, and those whose bulbs lasted too long were fined. Isn’t that crazy? Fining a business because their product works too well! The historian Markus Krajewski laid out the archival record in IEEE Spectrum in 2014. It is the cleanest case on file of an industry agreeing to make its product worse so that customers would buy it more often.

Cars are the example most people use. Now (at the time of writing this article in Q4 of 2026) ironically, if cars were being built to fail sooner, the cars on the road SHOULD be getting younger. They are older than they have ever been. S&P Global Mobility put the average age of a light vehicle in the United States at a record 12.8 years in 2025, and the European manufacturers’ association ACEA puts the average EU car at 12.7 years. But we need to consider that part of that is due to also decoumented price of new cars, electrification of vehicles, and so people have been in a way drivn to (pardon the pun) keep their old car longer when replacing it costs more. Even so, a fleet cannot reach an average age of nearly thirteen years if the machines are falling apart at six, right?

The stronger evidence in vehicles and equipment though concerns who is allowed to fix them. In January 2025 the US Federal Trade Commission sued Deere & Company over repair restrictions on farm equipment. The case settled in July 2026. For ten years, Deere must give farmers and independent repair shops the same repair resources it gives its own dealers, including the software needed to read fault codes and pair new parts. Deere did not admit wrongdoing.

The same pattern shows up in smaller products, and in each case there was a reason before there was an overreach:

Printers. A firmware update that HP sent to printers in November 2020 enabled a feature called Dynamic Security, which stopped cartridges from other manufacturers working. A lawsuit followed, and under the settlement approved in 2025, owners of certain models can decline those updates. HP paid no damages and admitted nothing.
Phones. In December 2016 Apple released an update that slowed iPhones with ageing batteries. The stated purpose was reasonable: it prevented unexpected shutdowns. Apple did not tell customers, many of whom bought a new phone when a new battery would have done. Thirty-three US states and the District of Columbia settled with Apple for $113 million, and a separate class action settled for up to $500 million.

The Apple case is worth pausing on because it shows how thin the line is. The engineering decision was defensible. The silence around it was what turned a fix into a sales driver.

When we owned what we bought

Software is where the loss of ownership is felt most sharply.

You paid once (for a license and 20 or more-digit keycode to your windows 95 OS), you received a copy, and that copy went on working for as long as the machine did. A PC from the 1990s, like the ones you see on those nostalgic PC boot-up videos, that has never touched the internet will still boot today and run everything that was installed on it. The seller had no way to reach it, change it, or switch it off.

In UsedSoft v Oracle in 2012, the EU Court of Justice held that when a rightholder provides a copy of a program and grants a licence of unlimited duration in return for a one-off fee, that transaction is a sale of the copy. The buyer may even resell it. Backup copies were protected as well. The EU software directive says a contract cannot prevent the user from making a backup copy needed for use, and US copyright law gives the owner of a copy the right to make an archival one.

The terms on paper barely moved but what has changed was the seller’s ability to enforce and revise them after the sale. The first large step was anti-copying technology. When Microsoft launched Windows XP in October 2001, it introduced product activation, which it described as a way to reduce “casual copying”, meaning one licence shared across several machines.

Thoe luck enough to have experience this in the early days would recall at university helping a friend or roommate out by copying the software. It was infringement, and it cost the people who wrote the software real money. Activation created a permanent line from the product back to the seller, and determined pirates removed the checks anyway. Paying customers were the ones who kept them.

What changed

If a single moment has to be named, the usual candidate is Adobe’s announcement in May 2013 that its Creative Suite would no longer be sold as a boxed, pay-once product and would be available only by subscription. I won’t hang the argument on one date, because the shift was gradual and more nuanced and reached different industries in different years. So I’ll focus on what conditions made this change rational, Yes, I said RATIONAL.

Products became permanently connected. You can blame the internet. Once a product is online for its whole life, the seller can update it, meter it, and switch features on or off after it has left the shop. That is true of software first and of anything with a chip in it later.

Running a service became cheap. When Amazon launched its S3 storage service in March 2006, a company no longer needed its own data centre to deliver software over the internet. Renting the infrastructure by the hour made it possible for a team of five to offer a service to the world.

Investors paid more for predictable income. A business that sells once has to win every sale again next year. A business with subscribers starts each year with most of its revenue already contracted. Markets price that difference heavily. McKinsey’s analysis of more than a hundred business software companies found the most highly valued quarter trading at a median of 24 times revenue against 5 times for the lowest quarter, and the measure most associated with the gap was how much revenue the company retained from its existing customers.

The switch drew a loud customer backlash, but did you know despite the protests revenue went from $4.06 billion in fiscal 2013 to $21.51 billion in fiscal 2024? So what do you think happens when a company successfully achieves growth in this way? Every with a profit target will want to understand it and copy it or tinker with it so works on their business model and strategy. The boards in the software industry could read those numbers and so they demanded ‘if they can do it, so can we’.

Put the three together and the picture is of companies responding sensibly to the conditions in front of them. A founder with a small team, real server costs and investors who value recurring revenue at several times the alternative would have needed a strong reason not to move to subscriptions.

Let’s be fair

It is easy to tell this as a story about what customers lost. But a fair account has to include what they were getting, because most of the reasons for the shift were good ones.

Connected software needs maintenance forever. That 1990s PC is safe only because it is offline. Anything connected needs security fixes for as long as it runs, and that work costs money every month. A one-off price cannot fund an open-ended obligation.

Storage really does fail. Hard drives and memory sticks die, and a photo library kept on one device is one accident away from being gone. Keeping copies in several places and replacing failed hardware is a real service with a real ongoing cost.

A product that lasts runs out of customers. A company that sells an excellent product once has to find new buyers forever or stop improving it. A smaller upfront price with ongoing payments can fund continuous development and lets a small team survive between major versions.

Subscriptions lowered the entry price. Professional tools that used to cost thousands up front became available for a monthly fee a freelancer could afford.

None of this is a trick. The customers who complain loudest about subscriptions frequently benefit from all four.

But is that fair?

The trouble is that every one of those reasons reaches a limit and businesses rarely stop there on their own.

The step	The fair reason	Where it overshoots
Activation and copy protection	Copying was real lost revenue	The checks end up constraining paying customers more than pirates
Updates and connectivity	Connected products need security fixes indefinitely	“Needs updates” turns into “stops working without our servers”
Subscriptions	One-off sales cannot fund ongoing work	Rent charged on something already delivered
Cloud storage	Devices fail and backup is a real service	Your own data becomes the thing keeping you subscribed
Controlled repair	Safety, quality and counterfeit parts are real concerns	Repair is blocked so that replacement is the only option

Examples of real cases.

In 2009, Amazon discovered that a publisher had sold copies of George Orwell’s 1984 through the Kindle store without holding the rights. Amazon removed the book from the store and also deleted it from the devices of customers who had bought it. The company apologised and said it would change its systems so that it would not happen again. The capability was the revealing part. A bookshop that sells you an unlicensed paperback cannot come into your house and take it back.

In December 2023, Sony told PlayStation users that television shows from Discovery that they had purchased would disappear from their libraries at the end of the month, because a licensing agreement had ended. After the reaction, the content was kept available under an extended arrangement. Customers had pressed a button marked “buy”.

BMW gave the clearest example in a physical product. It fitted heated seats to cars and then offered to switch them on for around $18 a month. The hardware was already in the car and already paid for. In September 2023 the company dropped the idea. Its board member for sales, Pieter Nota, put the lesson plainly: “It’s either in or out.”

In each case there is no recurring cost behind the recurring charge or the retained control. The seat heater does not need a server. A book already on your device does not need ongoing permission.

The low cost goods, physical enshittification

A two-euro item that breaks in a week is a poor product; so it is honestly worth €2. Nobody was promised quality so the cost sits somewhere in that low price. According to the European Commission’s figures, 4.6 billion low-value items entered the EU in 2024, about 12 million parcels a day, roughly double the year before, and 91% of them came from China. Each one was shipped individually across the world, and a large share will be thrown away quickly.

That is a problem of costs landing outside the transaction, which we covered in Greener Per Unit, Browner In Total. The remedy is different too. The EU has ended the duty exemption for low-value parcels and applies a charge of €3 per item from July 2026. It is a pricing fix for a pricing problem.

Ownership vs Subscription/Renting

The instinct behind the whole complaint is one I hold strongly: owning beats renting.

Housing is the obvious comparison. Pay a mortgage for twenty-five years and you have an asset. Even if the building loses value, the land remains, and there is something to sell or to pass on. Pay rent for twenty-five years and you have had somewhere to live, and nothing more.

Subscriptions share one half of that. Nothing accumulates. Ten years of payments for a software suite, a music catalogue and a photo archive leave you with pretty much nothing the moment you stop subscribing.

Now the bit that diverges from the house analogy; a house holds value because land is scarc, but A ten-year-old copy of a program is worth close to nothing on any market, however legally you own it. So, nobody builds wealth from owned software.

What ownership gives you with a digital product is control and permanence. The thing cannot be changed under you, repriced, or withdrawn. You decide when to replace it. For a business, that is the difference between a tool and a dependency. For a household, it is the difference between a photo album on a shelf and one that exists for as long as a card payment goes through.

Do this though for Pictures and music: just put it in a flashdrive, SSD, or on your own airgapped network. It’ll be yours even in its digital form; the cloud storage is stilll useful for the access from anywhere.

The test: can you leave?

All of this points to one question that works across every category. Can the customer leave, and what do they keep when they do?

It breaks into three parts:

Is there a recurring cost or recurring value behind the recurring charge? Security updates, storage and new features qualify. A heated seat does not.
Does the product keep working if you stop paying, or if the company changes its mind? At minimum, at the version you paid for.
Can you take what is yours? Your files, your photos, your records, in a form another product can read.

The test is useful because it disciplines both sides of the argument.

To the critics: a great deal of online anger is aimed at companies whose products people could drop tomorrow and choose to keep using. Where leaving is easy, continued use is a genuine vote. A company that earns its money that way has done exactly what a business is supposed to do, and resenting its success is just noise.

To the companies: where leaving is hard, continued use proves very little. People stay on a social network because their family is there. They keep paying for storage because fifteen years of photographs are in it. Usage figures in a locked-in market cannot tell you whether customers are satisfied or simply stuck, and a business that reads those figures as approval is going to be surprised when an alternative or a regulator arrives. Doctorow’s original point was exactly this: a product can be made worse without losing its users only when those users have nowhere to go.

So the same large company can be on both sides of the line at once, fairly earning loyalty with one product and coasting on switching costs with another. That is a more accurate picture than “big tech is evil” or “they are successful, so they must be right”.

Companies that survive without holding anyone hostage

The original question was what a company should do if it needs continuous revenue to survive. There are working answers, and none of them requires going back to the past.

Subscription with a floor. JetBrains, which makes software development tools, sells subscriptions. After twelve consecutive months of payment, the customer receives what the company calls a perpetual fallback licence. If they cancel, they keep a specific version permanently, without further updates. The company gets recurring revenue, and it has to keep earning it, because the customer can stop at any time and still have a working tool.

Charge for the result, and keep the risk. In 1962 Rolls-Royce began offering engine support for a fixed fee per flying hour under the name Power-by-the-Hour. The operator pays only while the engine is working, so every breakdown costs the manufacturer. Michelin does something similar for truck fleets, pricing tyres per kilometre driven. These are rental models, and they improve quality, because the maker now pays for every failure. The lesson for the ownership debate is uncomfortable but useful: what protects the customer is who bears the cost when the product fails, and a rental contract can arrange that as well as a sale can.

Sell durability at a price that reflects it. Miele tests its washing machines against a simulated twenty years of household use, around 5,000 wash cycles, and prices accordingly. The company is careful to say that this is a testing standard and not a guarantee. It is a smaller market than the mass one, and it has sustained the business for a long time.

Make money on repair and upgrades. A product designed to be opened, with parts and manuals available, turns maintenance into an honest revenue line. The customer returns because the product is worth keeping.

All four share one feature. The customer could walk away with something that still works, and the company makes money anyway.

Where the rules are heading

There is a practical reason to design for exit now. The law is moving that way, unevenly.

In the European Union:

Repair. The Right to Repair Directive has applied since 31 July 2026. For product groups including washing machines, fridges, vacuum cleaners and smartphones, manufacturers must repair on request at a reasonable price, even outside the guarantee period, and cannot use contract terms or software to block independent repair.
Batteries. From 18 February 2027, portable batteries in products sold in the EU must be removable and replaceable by the end user.
Data and switching. The Data Act has applied since 12 September 2025. It gives users the right to the data their connected products generate and a right to switch between cloud providers, with switching charges to be eliminated entirely from 12 January 2027.

In the United States, California has required since 1 January 2025 that sellers of digital goods stop using words like “buy” and “purchase” unless they disclose clearly that the customer is receiving a licence.

The direction is not uniform, and it would be misleading to pretend otherwise. A US federal rule that would have required subscriptions to be as easy to cancel as to start was struck down by an appeals court in July 2025 on procedural grounds. In Europe, a citizens’ initiative asking that video games remain playable after publishers end support collected 1,294,188 verified signatures. The Commission replied in June 2026 that it could not propose a legal obligation, and offered to broker an industry code of conduct instead.

Even so, the general movement is clear enough to plan around. A company that builds repairability, data export and a working fallback into its product today is choosing its own design. One that waits will have the design chosen for it.

The next round is already here

The same pattern is forming around artificial intelligence, in two opposite directions.

AI services are the purest rental so far. A business that builds its daily work around a model it accesses by subscription depends on something that can be repriced, changed or retired by its provider, and there is no old version in a drawer to fall back on. The three questions above apply to it directly, and anyone adopting these tools at scale should ask them before the workflows are built.

At the same time, AI is making it far cheaper to build software. A tool that once justified a monthly fee because nobody could afford to write their own is now something a small team can produce in days. That weakens the hold of any vendor whose main advantage was the cost of the alternative. Subscriptions that deliver real ongoing value will be fine. Subscriptions that survive mainly because leaving is painful are about to face customers with options.

Where this leaves the argument

Companies need to survive, and continuous revenue is a legitimate way to do it. Customers need to be able to leave, and to keep what they paid for when they do. Those two needs were never in conflict. The businesses that hold up best over the next decade will be the ones that would keep their customers even with the door wide open, and it is worth finding out which kind you are before someone else tests it for you.

Frequently Asked Questions

What does enshittification actually mean?: Cory Doctorow coined it in January 2023 to describe how platforms decay: good to users first, then squeezing users for business customers, then squeezing both. In everyday use it now covers three different things: platform decay, products that wear out or resist repair, and ownership replaced by licences and subscriptions.
Why did so many products become subscriptions?: Three conditions changed together. Products became permanently connected, cloud infrastructure made services cheap to run, and investors began valuing predictable recurring revenue far above one-off sales. Under those conditions, subscriptions were the rational choice for most businesses.
Is planned obsolescence real or a conspiracy theory?: Both exist. The Phoebus lightbulb cartel of 1924 is documented history. The belief that modern cars are built to fail has weaker support, since the average car on the road is older than ever. The better-evidenced modern problem is restricted repair.
Did people ever really own their software?: In practice, yes. The paperwork always said licence, but a one-off payment for a copy the seller could not reach worked as ownership, and the EU Court of Justice treated it as a sale in UsedSoft v Oracle. What changed was connectivity, which let sellers enforce and revise terms after the sale.
How can you tell a fair subscription from an unfair one?: Ask whether there is a recurring cost behind the recurring charge, whether the product keeps working at the version you paid for if you stop, and whether you can take your own data with you. A subscription that fails those is charging for the difficulty of leaving.

Äctvli Responsible Consulting helps organisations understand what they depend on, from suppliers to software, and what it would take to change course. If the three questions in this article were uncomfortable to answer about your own vendors, or about your own product, that is a conversation worth having.

Get in touch if you want to talk it through.

Sources

The term

Cory Doctorow, Pluralistic – Tiktok’s enshittification, 21 January 2023

Planned obsolescence and repair

Markus Krajewski, IEEE Spectrum – The Great Lightbulb Conspiracy, 2014
S&P Global Mobility – U.S. Vehicle Age Rises Again to 12.8 Years in 2025, 21 May 2025
ACEA – Vehicles on European Roads, January 2026
US Federal Trade Commission – FTC, States Secure Settlement with Deere & Company, 8 July 2026
The Register – HP Inc settles printer toner lockout lawsuit, 19 March 2025
Texas Attorney General – $113 Million Settlement with Apple Over iPhone Throttling, November 2020

Software ownership

The shift to subscriptions

Where it overshoots

NPR – Amazon’s ‘1984’ Deletion From Kindle Examined, 24 July 2009
TechRadar – PlayStation will soon remove Discovery content from your library, even if you’ve paid for it, December 2023
TechCrunch – BMW feels the heat, stops charging for warming cheeks, 7 September 2023

Low-value imports

European Parliament – EU targets low-value imports via e-commerce platforms, 2025

Models that work

Regulation

European Commission – Directive on repair of goods
European Commission – Guidelines on the removability and replaceability of batteries, C/2025/214
European Commission – Data Act
Greenberg Traurig – Cloud Switching Under the EU Data Act, September 2025
Sidley – California’s New Digital Goods Law AB 2426, November 2024
Cooley – Eighth Circuit Vacates Entirety of FTC’s Negative Option Rule, 11 July 2025
European Citizens’ Initiative – Stop Destroying Videogames, Commission reply of 16 June 2026