Category: The Enshittification Machine: Platforms That Rot After Winning

A study of what happens after a platform wins.

This category follows the familiar cycle: attract users with a useful service, attract businesses with access to those users, then gradually squeeze both sides once leaving becomes inconvenient. Features disappear, ads multiply, search gets worse, subscriptions creep in, moderation becomes inconsistent, interfaces grow hostile, and the original product gets buried under monetization machinery.

The focus is on platform decay as a system rather than a single bad redesign. Network effects, lock-in, investor pressure, acquisitions, algorithmic control, subscription creep, API restrictions, creator dependence, and the slow conversion of useful public spaces into extraction machines all belong here.

The machine usually works beautifully right up until you notice what it is producing.

Posted on by

Measuring platform decline separately from personal nostalgia

“This site used to be better” is one of the Internet’s oldest recurring posts.

Sometimes it is true.

Sometimes the user was nineteen.

Those are not the same diagnosis.

Nostalgia is a terrible benchmark because both the service and the person changed. A platform may genuinely become more expensive or restrictive while also gaining useful features the old version never had.

The way out is measurement.

Dropbox gives us a change that nostalgia cannot explain away

Dropbox’s current documentation says free Basic accounts can be logged into on no more than three devices at a time.

The same help page contains the historical clue that matters: devices connected before March 2019 were grandfathered even when a Basic account already had more than three. Adding a new device requires disconnecting another or upgrading. See Dropbox’s device-limit documentation.

That is an objective product change.

A free user who synchronized five computers and phones before March 2019 could keep those existing links, but could not later replace or add devices without falling under the three-device ceiling.

You do not need to remember the old Dropbox homepage correctly to measure that.

Decline needs a defined dimension

A platform can improve in one way and decline in another.

Useful before-and-after measures include:

  • subscription price,
  • storage or device limits,
  • number of ads per task,
  • percentage of feed occupied by recommended content,
  • API capabilities,
  • export completeness,
  • response time,
  • outage frequency,
  • moderation turnaround,
  • number of steps required to perform a common action,
  • features moved from free to paid access.

If one of those gets worse, say that got worse.

Do not automatically promote the finding into the entire platform is ruined.

Personal behavior can create a fake decline too

A social network may feel boring because your original friends stopped posting.

A game community may feel empty because you changed games.

A music service may appear repetitive because your listening narrowed.

A search engine may feel worse because the kinds of queries you make changed.

Those experiences are real, but they do not isolate a platform change.

A better investigation looks for old documentation, archived screenshots, price tables, terms, changelogs, support pages, independent measurements, or user-interface captures that establish a before-and-after difference.

Enshittification should be falsifiable

If every memory of a better Internet automatically counts as evidence, the theory becomes useless.

A serious claim should be able to survive a question like:

What exactly changed?

Dropbox Basic’s three-device limit is a good example because the company itself documents the March 2019 boundary.

Maybe a user still prefers modern Dropbox.

Maybe the service is faster, safer, or more capable in other respects.

None of that changes the narrower fact that one free-plan capability became more restricted.

That is how platform decline should be studied:

one measurable deterioration at a time, with nostalgia kept outside the lab.

Posted on by

The difficulty of sustaining a service funded by permanent user growth

“Get huge first. Figure out the business later.” is a strategy with an expiration date.

Growth can hide that expiration for a long time.

A digital service can subsidize free users while investors, paid customers, advertising, or other revenue covers the gap. As long as adoption keeps accelerating, the growing user base itself can look like proof that the model is working.

Eventually the service still has to pay the bill.

Servers do not accept monthly active users as currency.

Evernote’s free tier shows the squeeze clearly

Evernote spent years using a substantial free plan as the front door to its note-taking ecosystem.

Then the free experience became much narrower.

In November 2023, Evernote announced that beginning December 4, Free accounts would be limited to 50 notes and one notebook. Existing users above those limits could still view, edit, export, share, and delete their material, but they could not keep creating indefinitely without reducing their data or upgrading. See Evernote’s 2023 Free-plan announcement.

In August 2024, Evernote added another restriction: Free users would be able to connect only one device at a time. See Evernote’s 2024 Free-plan update.

Evernote explained the economics directly. It said subscriber support funded infrastructure work, synchronization improvements, AI features, stability work, security, and continued development.

That does not make the restrictions pleasant.

It does make the underlying problem real.

Free growth is not the same thing as sustainable revenue

A generous free tier can help a service:

  • grow quickly,
  • create word of mouth,
  • establish habits,
  • become a default recommendation,
  • build network or file-history switching costs,
  • and produce a large pool of potential future customers.

But every non-paying user can still create costs: storage, synchronization, bandwidth, abuse handling, support, security, and engineering complexity.

If the business depends forever on tomorrow’s growth making today’s economics worthwhile, eventually tomorrow arrives.

There are several exits from the growth treadmill

A mature service can try to support itself through:

  • subscriptions,
  • advertising,
  • transaction fees,
  • enterprise contracts,
  • premium features,
  • donations or memberships,
  • lower operating costs,
  • or a deliberately smaller scope.

Each model changes the incentives differently.

The Enshittification problem begins when the transition to sustainability extracts disproportionate value from users who already built dependence under more generous terms.

That is why Evernote is useful as a documented example without requiring a conspiracy theory.

The company openly tightened the free tier and openly argued that paying customers support ongoing development.

The lesson is more general:

User growth can postpone the funding question. It cannot answer it forever.

Posted on by

Investor growth targets and pressure to increase revenue per user

A platform can run out of new humans before it runs out of revenue targets.

That is when revenue per user becomes interesting.

Public companies routinely measure how effectively an existing audience is being monetized. The metric is not sinister. It is useful accounting.

It also creates a very clear incentive:

If the audience is already large, growth can come from earning more from each person already there.

Reddit says this part out loud because investors need the numbers

Reddit defines average revenue per unique, or ARPU, as quarterly revenue divided by average daily active uniques in a geography. In its 2025 annual report, Reddit described multiple levers for expanding monetization and ARPU, including advertising products, advertiser growth, and other revenue opportunities. See Reddit’s 2025 Form 10-K.

The numbers moved substantially.

For the fourth quarter of 2025, Reddit reported global ARPU of $5.98, up 42% from $4.21 a year earlier. Daily active uniques rose 19% over the same period. See Reddit’s Q4 and full-year 2025 results.

That is not evidence that Reddit became worse.

It is evidence that the company became much better at producing revenue from the audience.

Those are different claims.

Monetization pressure can translate into product pressure

A platform trying to increase revenue per user has several obvious options:

  • show more advertising inventory,
  • improve ad targeting or pricing,
  • add premium subscriptions,
  • move features into paid tiers,
  • sell promoted placement,
  • license data or content,
  • charge businesses for tools,
  • increase marketplace commissions.

Some of those changes can improve the product.

A successful subscription may fund better moderation or infrastructure. Better ad tools can reduce irrelevant advertising. Business products can subsidize free users.

The problem appears when the revenue objective begins consuming more user value than the platform creates in return.

The metric does not prove the motive of every feature

This is where Enshittification analysis needs discipline.

If a company reports rising ARPU and later adds a disliked feature, that does not automatically prove the feature was ordered by investors to squeeze users.

Stronger evidence comes from investor materials that explicitly connect monetization goals to product strategy, followed by documented changes in pricing, ad inventory, fees, or paid access.

Reddit’s filings are useful because the company itself discusses monetization and ARPU as strategic objectives.

The larger mechanism is not unique to Reddit.

A platform’s early question is often:

How do we get everybody here?

Once the answer is sufficiently successful, the financial question naturally becomes:

How much can this audience produce?

That second question is where the user’s experience and the investor’s spreadsheet begin sharing the same screen.

Posted on by

Platform acquisitions that remove meaningful alternatives

Buying a competitor does not automatically kill the competitor.

Sometimes the acquired product grows.

Sometimes it stays independent.

Sometimes it becomes a feature inside the buyer’s platform.

And sometimes it disappears.

Wunderlist provides a clean example of the last path.

Microsoft bought Wunderlist, then retired it

Microsoft announced its acquisition of 6Wunderkinder, the company behind Wunderlist, on June 2, 2015. At the time, Microsoft praised Wunderlist as a market-leading to-do app known for simplicity, design, and broad availability across mobile, desktop, and web platforms. See Microsoft’s 2015 acquisition announcement.

Wunderlist did not disappear immediately.

Microsoft continued operating it while developing Microsoft To Do and providing migration tools.

Then the independent product ended.

Microsoft’s lifecycle record lists May 7, 2020 as the end of support for Wunderlist and directs users to migration guidance. See Microsoft’s Wunderlist lifecycle page.

That produces a straightforward before-and-after record:

  • 2015: Microsoft acquires a distinct cross-platform task service.
  • 2020: Wunderlist support ends and users are directed toward Microsoft’s replacement.

The user loses an alternative even when migration exists

A migration tool can preserve tasks.

It cannot preserve the existence of the old competitor.

Before the shutdown, a user could choose between Microsoft products and Wunderlist as a separately branded service with its own design decisions and roadmap.

Afterward, that particular choice no longer existed.

This does not mean Microsoft eliminated all competition in task management. Todoist, Things, TickTick, Apple Reminders, Google Tasks, and many others remained.

That distinction matters.

The claim should be proportional to the evidence.

Microsoft removed one meaningful alternative, not the entire market.

Acquisitions need outcome-based analysis

A lazy version of the Enshittification argument says:

Big company bought smaller company. Therefore bad.

That is not enough.

A better audit asks:

  • Did the acquired product remain independently available?
  • Were its terms or pricing changed?
  • Was its technology folded into the parent company’s product?
  • Could users continue using it without migrating?
  • Did the acquisition reduce the number of genuinely distinct choices?

Wunderlist answers those questions unusually clearly because the product has a documented acquisition date and a documented end date.

The important event is not the purchase by itself.

It is what happened to the alternative afterward.

Posted on by

Businesses dependent on a marketplace they do not control

A marketplace can be the reason a small business exists.

That is both its value and its leverage.

A seller does not join only for shopping-cart software. The platform supplies buyers, search traffic, payments, reputation systems, fraud handling, reviews, advertising tools, and a familiar checkout process customers already trust.

Rebuilding all of that independently can be much harder than opening a new website.

Etsy shows how a fee change lands after scale exists

Etsy announced in February 2022 that its transaction fee would rise from 5% to 6.5%, effective April 11, 2022. The company said it expected to reinvest most of the incremental revenue in marketing, seller tools, support, and marketplace quality. See Etsy’s fee-change announcement and its 2022 annual report.

That was not Etsy’s first fee increase. In 2018, Etsy raised its transaction fee to 5%, describing it as its first-ever fee change and tying the increase to larger marketing and product investments. See Etsy’s 2018 announcement.

By 2022, Etsy said its marketplace connected sellers with a built-in audience of more than 90 million shoppers.

That audience is the important part.

The seller cannot move the marketplace with the shop

A merchant can copy product photos to Shopify, WooCommerce, or another marketplace.

What does not automatically follow is:

  • the buyer traffic,
  • accumulated reviews,
  • marketplace search ranking,
  • customer trust in the checkout system,
  • saved favorites,
  • platform advertising data,
  • or the habit of shoppers starting their search on Etsy.

The seller may dislike a fee increase and still decide that leaving costs more than paying it.

That does not prove the fee is unfair.

It proves dependence changes the negotiation.

Marketplace power is partly optional and partly structural

Etsy’s position was that higher fees funded services intended to help sellers sell more. A seller could reasonably conclude that the exchange remained worthwhile.

Another seller could conclude the extra percentage materially reduced already-thin margins.

Both can be true.

The Enshittification question is not simply did the platform charge more?

It is:

How much practical bargaining power does a merchant retain after the platform becomes the source of customers?

A marketplace is most attractive when it can deliver demand a small business cannot cheaply create alone.

That same advantage is what makes walking away difficult when the platform later changes the price of access.

Posted on by

Creators dependent on audiences they cannot contact independently

A million followers are not the same thing as a million email addresses.

That difference is easy to miss while the platform is working well.

A creator publishes. Followers see the post. Notifications arrive. Recommendations send new viewers. The audience number grows and begins to look like an asset owned by the creator.

Then the platform changes how delivery works.

The creator discovers what the number actually represented.

A subscription is still a platform relationship

YouTube’s current documentation makes this unusually clear.

Subscribing to a channel puts new uploads into the viewer’s Subscriptions feed, but notifications are controlled by additional settings. By default, viewers receive personalized notifications rather than guaranteed alerts for every upload. Device settings can block notifications, and YouTube limits channels to three upload, premiere, or livestream notifications in a 24-hour period. See YouTube’s subscriber notification guidance and notification troubleshooting page.

The creator does not receive a portable directory containing the private email addresses or phone numbers of subscribers.

YouTube mediates the relationship.

That is normal for a platform.

It is also dependence.

Delivery rules can change after the audience exists

In August 2020, YouTube stopped sending the old new-upload email notifications that some subscribers had used as a direct-looking delivery channel for years. The change was announced through TeamYouTube support materials and immediately produced complaints from users who had organized their viewing around those emails. Current YouTube notification documentation centers delivery on the platform’s own notification systems and viewer settings.

The creator’s subscribers did not vanish.

The route between creator and subscriber changed.

That is the important mechanism.

A follower count can remain numerically identical while practical reach changes because the platform modifies:

  • feed ranking,
  • notification defaults,
  • daily notification limits,
  • recommendation systems,
  • account eligibility,
  • or moderation and monetization rules.

Owned audiences behave differently

An email list, RSS subscriber base, or self-hosted membership system has its own risks and dependencies, but the creator normally has a more direct relationship with the recipient.

A platform follower is different.

The creator has permission to publish inside the platform.

The platform decides much of the delivery.

That is why creators often ask followers to join newsletters, websites, or memberships outside the social service. They are trying to convert platform reach into a relationship that survives platform changes.

The Enshittification Machine becomes powerful when a creator’s livelihood depends on an audience number that looks owned but behaves rented.

The creator may have earned the attention.

The platform still owns the hallway between them.

Posted on by

Contact graphs that cannot move cleanly between services

You can export a phone number.

You cannot export the fact that twelve specific people are already in the family group chat and actively use it every day.

That is the portability gap between contacts and relationships.

A contact list is a set of identifiers.

A contact graph includes who follows whom, which groups exist, who belongs to them, who has blocked whom, which conversations are active, which memberships are mutual, and which people actually still use the service.

Those things are much harder to move.

Data portability does not automatically recreate a network

The European Union’s Digital Markets Act now requires designated gatekeepers to provide stronger portability tools for user-generated data. The Commission describes Article 6(9) as requiring free tools and continuous access so users can move data to authorized third parties. See the Commission’s end-user data portability guidance.

That is a significant improvement over a platform merely offering a giant archive download.

But even a perfect export cannot force the people in the file to create accounts somewhere else.

A social graph is partly data and partly coordinated human behavior.

Messaging interoperability shows another approach

The DMA treats messaging differently because strong network effects make migration unusually difficult. Instead of requiring every user to move, Article 7 requires certain gatekeeper messaging services to interoperate with other providers.

The European Commission says the first third-party messaging services interoperable with WhatsApp launched in November 2025. The rules are being phased in across one-to-one messaging, group chats, and eventually calls. See the Commission’s messaging interoperability page.

That is conceptually important.

If two services can communicate, a user can leave one interface without necessarily abandoning everyone still using it.

The solution attacks the network lock-in directly.

Exportable is not the same as portable

A platform can truthfully say:

You can download your data.

That may still leave the user with:

  • a CSV of contacts who are not on the new service,
  • old messages with no live destination,
  • groups that must be rebuilt manually,
  • followers who must choose to follow again,
  • community roles that do not transfer,
  • and years of relationships represented only as static records.

The file moved.

The network did not.

That distinction matters whenever a dominant platform points to an export tool as proof that users are free to leave.

Freedom to leave is not measured only by whether a button produces a ZIP archive.

It is measured by how much useful social structure survives the trip.

Posted on by

Switching costs created by accumulated personal history

A file is easy to move.

Ten years of context is not.

That difference is one of the strongest forms of platform lock-in.

People do not merely accumulate bytes inside long-lived services. They accumulate albums, comments, tags, favorites, search history, shared links, organization systems, social relationships, memories, and habits built around where those files live.

The longer the history, the less useful the phrase just export your data becomes.

Flickr shows how a generous archive can become a switching problem

In 2013, Flickr announced that every member would receive 1 terabyte of storage for free, enough by its estimate for more than 500,000 photos. See Flickr’s official 2013 account-limits announcement.

For years, that made Flickr a plausible home for enormous personal photo histories.

Then the economic terms changed.

In 2018, Flickr announced that free accounts would be limited to 1,000 photos or videos. Members above the limit were told to upgrade to Pro or download material beyond the new allowance. Flickr later extended the deadline for affected accounts. See Flickr’s official Free and Pro plan announcement and timeline extension.

That is a clear before-and-after change.

It also shows why switching cost is not the same thing as file size.

Downloading the photos is not the same as moving the history

A user can copy image files to a hard drive.

What does not move as neatly is the surrounding structure:

  • albums and ordering,
  • comments from other people,
  • favorites and engagement,
  • links from old websites,
  • group memberships,
  • captions and tags,
  • search visibility,
  • relationships with other photographers,
  • years of muscle memory about where everything is.

Some of that can be exported or rebuilt.

Some cannot.

A rival service can offer better pricing tomorrow and still be less useful than the worse service containing fifteen years of your past.

History creates dependence without a contract

Nobody has to sign a ten-year lock-in agreement.

The archive itself performs the locking.

This matters because platform operators can change prices, limits, interfaces, moderation rules, or feature availability after users have accumulated much more inside the service than they had when they first joined.

That does not mean every later restriction is illegitimate. Flickr’s owners openly argued that the old free-storage model was not sustainable and that Pro subscriptions were needed to support the service.

The useful measurement is simpler:

How much more expensive is leaving now than it was when the user first arrived?

If the answer includes years of personal history rather than merely a ZIP file, the platform has acquired leverage that did not exist on day one.

Posted on by

Network effects that keep dissatisfied users from leaving

A social network is not valuable only because its software is good.

It is valuable because your people are already there.

That makes leaving a different decision from replacing a calculator app.

A better calculator can win one user at a time.

A better social platform may need to win your family, coworkers, hobby group, local businesses, school parents, favorite creators, and everybody else you actually came to talk to.

That is a network effect.

The crowd becomes part of the product

The UK’s Competition and Markets Authority described this problem directly in its 2020 study of online platforms and digital advertising. It found strong network effects in social media and noted that Facebook had by far the largest network. Other services were often used alongside Facebook rather than as complete substitutes for it. See the CMA’s market-study presentation.

That explains an otherwise strange behavior.

A person can complain about a service every week and still open it every morning.

The contradiction disappears once the user is not choosing only between products.

They are choosing between staying where the network exists and leaving without the network.

Dissatisfaction does not automatically create an exit

Suppose a messaging app makes an unpopular policy change.

One user can install a rival in thirty seconds.

That does not move the family chat.

If Grandma, the soccer team, three clients, and the neighborhood group remain on the old platform, the dissatisfied user may end up maintaining both apps. The rival becomes an addition instead of a replacement.

The European Union’s Digital Markets Act explicitly recognizes this problem for messaging. Its interoperability rules say large messaging gatekeepers benefit from strong network effects and require certain services to interoperate with competitors. The European Commission says the first third-party messaging services interoperable with WhatsApp launched in November 2025. See the Commission’s messaging interoperability overview.

That is an important structural change.

Instead of requiring everybody to migrate together, interoperability can let users on different services continue communicating.

The alternative has to preserve the relationship, not merely the feature

A competing platform can have:

  • cleaner software,
  • fewer ads,
  • better moderation,
  • stronger privacy,
  • lower prices,
  • and a nicer logo.

None of that solves the network problem if the people a user needs remain somewhere else.

This is one reason platform decline can continue longer than ordinary product logic suggests.

A company does not necessarily lose users the moment it makes the service worse.

The accumulated network can absorb some deterioration.

That does not prove every dominant platform is deliberately abusing lock-in.

It does explain why user dissatisfaction and user departure are different measurements.

In the Enshittification Machine, the crowd is not merely using the platform.

The crowd is part of what makes the platform hard to escape.

Posted on by

The withdrawal of introductory subsidies after a platform establishes dependence

The easiest time to offer an incredible deal is before everybody depends on it.

Google Photos provides a clean example.

When Google launched the standalone Photos service in May 2015, one of its headline features was free unlimited storage for photos and videos uploaded under its High Quality setting. Google described the product as a new home for people’s photo libraries and promoted unlimited High Quality storage as a reason users would no longer have to worry about where to keep their snapshots. See Google’s 2015 launch announcement and its later 2015 product recap.

For years, that was an unusually generous subsidy.

Then the terms changed.

The unlimited offer ended for new uploads

In November 2020, Google announced that beginning June 1, 2021, new photos and videos uploaded in High Quality would count against the 15 GB of storage included with a Google Account or against additional storage purchased through Google One. Existing High Quality uploads made before the deadline were grandfathered and did not suddenly consume the quota. See Google’s storage-policy announcement.

Google said the change was necessary to keep up with growing demand and build the service for the future. At the time, it said Google Photos held more than 4 trillion photos and was receiving 28 billion new photos and videos each week.

Those facts matter because they prevent a lazy conclusion.

This was not evidence that Google secretly planned in 2015 to trap users and charge them later.

It is a documented example of a highly subsidized platform benefit being reduced after enormous adoption.

Dependence changes the practical effect of a price change

A storage service is not like a coupon for one sandwich.

Its usefulness accumulates.

People upload years of family photos. Automatic backup becomes a habit. Search indexes the library. Albums and shared links are created. Phones are configured around the service. Other local copies may be neglected because the cloud copy feels permanent.

By the time the economic terms change, the user is not evaluating a fresh product from zero.

They are evaluating the cost of moving an existing archive and changing an established workflow.

That is platform dependence.

Not every subsidy withdrawal is enshittification

A company can legitimately discover that an unlimited offer is economically unsustainable.

Storage costs money. Demand grows. Products change.

A responsible analysis therefore separates three questions:

  1. What were the original terms?
  2. What changed, and when?
  3. What switching costs had users accumulated by then?

The answer should come from documented terms and practical effects, not from pretending every price increase proves bad faith.

That distinction is important for the new section, The Enshittification Machine — Platforms That Rot After Winning.

The interesting mechanism is not simply that a service gets more expensive.

It is that platforms can spend aggressively to make adoption irresistible while users are still free to choose, then recover more value later when leaving has become harder.

The subsidy helps build the dependence.

The dependence changes what happens when the subsidy disappears.