Sheftali's Unknown

The standard advice is to cancel whatever you have stopped watching. Try it on a streaming service and watch where it breaks: you click cancel, and instead of leaving you are handed a cheaper plan, a free month, or a tier you did not know existed. Streaming retention offers are the reason a decision to leave so often ends with you still subscribed, paying less, and watching a service that now earns more from you than before.

What Are Streaming Retention Offers?

They are discounts, free months or a cheaper advertising tier presented at the moment you try to leave, rather than at renewal when you are deciding whether the service is worth its price at all.

Ampere Analysis published a three year pricing dataset in August 2026 covering Netflix, Disney+ and Amazon. The average annual price increase fell from 24 per cent in 2023/24 to 14 per cent in 2025/26. Two years, and the size of the ask nearly halved. Most coverage read that as competition finally working.

I would argue the opposite. Nothing about this market got more competitive. Households started cancelling, and a cancellation is worth far less to a streamer than a downgrade, so the industry built a downgrade worth having and made it the cheap answer to a question it asks you at the door. The price stops being a number on a rate card and becomes a function of what you will tolerate, which is the same mechanism behind the way your own data quietly sets the price you are shown.

Why The Cheaper Tier Only Appears At The Exit

Because a customer who never asks is worth the full price, and one who reaches the cancel screen has just told the company exactly how much they were about to stop paying, which is the only number a retention offer has to beat.

Advertising is what makes the offer affordable. Ad inventory scales with viewing hours rather than billing events, so a household that downgrades and keeps watching can be worth more than one paying more and watching less. Netflix advertising was tracking near three billion dollars during 2026, roughly double the previous year. Charging for extra member slots after the password sharing crackdown adds a second meter, pointed at people who were already inside the account.

Jaanika Juntson of Ampere Analysis put the mechanism plainly when the data landed: advertising has become an increasingly important revenue stream that reduces reliance on subscription pricing alone, while password sharing crackdowns let streamers get more value out of the audience they already have. Read that twice. None of it is about giving anybody a better deal, and it is the same quiet transfer as hardware you already paid for turning into a subscription.

Does Downgrading Actually Save You Money?

Sometimes, and the way to tell is to price the gap rather than the plan, because the number that moved over the past three years is the distance between the two tiers rather than either tier on its own.

Work down that middle column and the shape separates from the headline. Every service level figure describes a slowing rate of increase on a plan price. Not one describes the thing that actually grew, which is the distance between the two plans you are asked to choose between. That distance is the product now, and it widens even in a year when neither headline price moves.

Region matters more than the averages suggest. Ampere put North American subscribers at $1.70 an increase, about 15 per cent of the prior price, Western Europe at $1.86 and 16 per cent, and Central and Eastern Europe at $1.68 and 18 per cent, the steepest proportion of any region measured. Same companies, same three years, and the share of your old bill being added varies by where you happen to live. None of that appears on a rate card, and none of it is something a subscriber gets to negotiate, which is worth holding in mind before reading any retention offer as generosity.

What We Are Not Able To Measure

Nobody publishes how many minutes of advertising an hour each tier carries, or how that has moved since these plans launched, which means the thing most likely to erode the value of a downgrade is the one nobody can put a number against.

I will flag that as my read rather than a finding. Ad load is not contractually fixed. A plan carrying four minutes an hour today can carry seven in eighteen months with no price change, no notification email and nothing a comparison site would record as an increase. An annual plan locks the price and does not lock what the price buys. We also do not cover Indian pricing on these services in this piece, because the tier structures differ enough that the dollar figures above would mislead. A claim you cannot check is the common thread here, the same reason a fake spare part goes unnoticed at a repair counter, and the reason a family plan can quietly cost more than it saves.

Open your account settings tonight and find what the ad-free option costs you above the advertising plan. If that gap is larger than you would have accepted as a straight price rise, you already took the increase. You took it as a choice, which is exactly why it was arranged that way. Start the cancellation flow before you decide anything, look at what appears, and then close the tab and think about it for a day. The offer will still be there tomorrow, because it was always available.

Your child spikes a fever at eleven at night. You open a retail app, type in baby thermometer, and the first page hands you the expensive ones. Not because they rank better. Because something in the profile attached to your account has decided you are a new parent, awake late, in no position to shop around. The Federal Trade Commission used almost exactly that scenario as its own illustration, which tells you it is not hypothetical.

Why It Matters

Start with the distinction the whole thing turns on. Dynamic pricing moves with supply, demand and inventory, and it hits everyone in the market at the same moment. Personalized pricing moves with you. The same product on the same page can carry a different number for you than for the person sitting next to you, because the retailer has built an estimate of what you specifically will tolerate. The first is economics. The second is a guess about your wallet, assembled from data you did not knowingly hand over for that purpose.

And here is the part most of the coverage underplays: the money is not the real injury. A few percent on a thermometer will not ruin anyone. The injury is that you cannot detect it. A shortage you can read about. A surge price announces itself in the checkout screen. A price built quietly from your browsing history looks exactly like an ordinary price tag, which is precisely why it works.

A January 2025 Federal Trade Commission study of six pricing intermediaries, Mastercard and McKinsey among them, documented what actually feeds these systems: precise location, browser and search history, cart abandonment, demographics, purchase history, and mouse movements on the page. Mouse movements. The hesitation before you click is an input. It is the same shape as a vendor rewriting the deal after you have already paid, the pattern behind cloud shutdowns bricking smart home devices you already own, except this one happens before the sale instead of after it.

Four figures explain the shape of the problem better than any argument about fairness does. One is a deadline. One measures who absorbs it hardest. One is scale. One is the reason the industry will fight for it.

Take the client reach figure. That is what separates this from a story about two or three technology giants running clever experiments on their own customers. These pricing engines are sold as a service, bolted onto grocery chains and apparel retailers that have never written a line of machine learning in their lives. Your local supermarket does not need a data science team to price you individually. It needs a vendor contract.

What The Proposal Actually Says

So what does it require, and where does it stop? Here is the whole of it, stripped of the compliance language the law firm client alerts are written in.

Read the safe harbour row twice. If a retailer prices you off your own purchase history on your own logged-in account, and says so plainly, it is compliant. That description covers a very large share of what grocery loyalty programmes already do today, which means the practice most shoppers would object to hardest is also the one most cleanly blessed.

Friction Points

The proposal does not ban any of this. It cannot, or rather the FTC says it cannot, reading its own authority as reaching disclosure and stopping there. Chairman Andrew Ferguson framed the expectation plainly: when consumers see a listed price, they expect it to be the same price that everyone else sees, not the retailer's estimate of how much they are willing to pay based on their personal data. That is an accurate description of what shoppers assume. It is not a description of what the policy delivers.

Here is where I part company with the optimistic reading. A disclosure tells you a price was personalized. It does not tell you whether yours is the high one. Without a baseline, without the unpersonalized number sitting beside it for comparison, the label is a weather warning with no temperature attached. My honest view is that this is the unsolved weak point, and the comment docket will not solve it either, because the obvious fix, publishing a reference price alongside the personal one, is the single thing retailers will refuse outright.

None of this is unusual for how consumer terms get rewritten on people quietly. It is the same posture as the way Windows 10 extended security update terms landed without an announcement, changed in place, visible only to whoever went looking. What you can control is your own measurement discipline. Four things are worth watching:

• Signed in versus signed out. If you only ever see the app's logged-in price, you have never seen the other one.
• Loyalty cards. The discount is real. So is the profile it builds while you collect it.
• Location permissions. Precise location sits on the FTC's list of pricing inputs, and most retail apps request it by default.
• Repeat visits. Cart abandonment is a documented input, so hesitating on a product page is itself a signal you are sending.

Do one thing this week. Pick something you buy on a schedule, check its price signed in on the app, then check it again in a private browser window with location switched off, and write both numbers down. If they match, good, you have a baseline. If they do not, you have learned more about how you are priced than any disclosure label is ever going to tell you. Start measuring now, while the comparison still surprises you.

Related: the auto-renewal law now requires clearer consent before your card is billed again

Your laptop is fine. It boots fast, the battery still holds a charge, and the only thing wrong with it is a processor that missed Microsoft's Windows 11 eligibility list by about two model years. Last autumn it stopped getting security updates anyway. Then in late June, with no keynote and no email, Microsoft handed it another year.

The Extra Year Nobody Announced

The extension arrived the way awkward decisions usually do, as an editor's note appended to documentation that was already sitting there. Microsoft's own consumer ESU page now prices the paid route at $30 one time plus tax, and dates the coverage window to 12 October 2027. Anyone who enrolled during the first year was carried across at no extra charge and told nothing. Call it a reprieve. Actually, call it a pressure valve, because that is what it does.

The pressure is real and it is measurable. Windows 11 holds a clear majority of desktops now, but a large minority of the world's PCs are still running the version Microsoft stopped patching for free. Those machines mostly do not belong to holdouts making a point. They belong to people whose hardware failed an eligibility check they had never heard of until a notification told them about it, which is the same pattern that switched off smart home devices their owners had already paid for. And the obvious fix is not cheap, because memory pricing has pushed the cost of a decent new machine well past where it sat two years ago.

Four numbers describe the entire offer. What Microsoft added, what the points route costs, how far a single license stretches, and how much of the world this still applies to.

The device count is the number people leave on the table. Enrollment feels like a per-machine chore, so it gets done on the laptop that nagged loudest and forgotten on the desktop in the study and the machine the kids use for homework. A single license stretches across a household's worth of PCs provided they all sign in under the same Microsoft account. That turns the free route into something worth doing deliberately, in one sitting, rather than reactively.

Which Door To Walk Through

The routes are not equivalent, even though three of them end in exactly the same patch stream. What separates them is what you hand over on the way in, and how easily you can change your mind later.

Read that as a privacy decision rather than a price decision, because on coverage the three ESU columns are indistinguishable. The paid route is the only one that buys patches without handing Microsoft either your synced configuration or your accumulated goodwill, which is a strange thing to have to pay for and, for some readers, still the right call.

Where This Still Bites

Extended Security Updates is not a support extension, and the naming does a lot of work to hide that. It is a patch drip aimed at one specific build. Microsoft states plainly that enrollment brings critical and important security updates and nothing further: no fixes for bugs that annoy you, no product improvements, no help desk to call when something breaks. If a driver stops working next March, you are on your own with it.

The larger risk sits outside Microsoft's control entirely. Browser makers, GPU driver teams, antivirus vendors and banking apps set their own end-of-support dates, and none of them are obliged to match this one. An operating system that receives patches while the software layered on top quietly stops being tested is only partly protected. Microsoft's household pricing has form for shifting terms mid-stream too, as family plan subscribers discovered when Copilot arrived in their bill. And if you are switching on Windows Backup purely to unlock enrollment, treat it as one layer of a backup plan that can actually survive a dead drive, not the whole of one.

Whether a third year appears is the genuinely open question, and nobody outside Redmond can answer it. My read is that the June extension was a response to adoption numbers that refused to move, which makes another one plausible but not something to bet a household's security on. Treat this as the last cheap year rather than the first of many. That is a stance, not a forecast, and I would happily be wrong about it.

• Apply the license across every PC in the house in one sitting, not one crisis at a time.
• Go looking for the enrollment prompt in Windows Update rather than waiting for it to find you.
• Treat the account you enroll with as permanent, because unpicking it later reopens the whole question.
• Put the decision date in your calendar for spring 2027, not autumn, since autumn reminders get snoozed.

Enroll this week, on every machine in the house, through whichever door you can live with. Then open your calendar and book a real decision for spring, because a year of borrowed patches is not a plan and the clock started in June without asking you.

The plug still clicks. The hardware in the wall works exactly as well as it did the day it shipped, the relay closes, the light comes on if you press the button with your thumb. What stopped working was permission. A server in someone else's building got switched off, and the thing you paid for quietly demoted itself to a switch you have to walk across the room to reach.

Why Your Own Hardware Stopped Answering

Belkin ended cloud services and app support for most Wemo products on 31 January 2026. Plugs, wall switches, bulbs, baby monitors, motion sensors: all still drawing power, none of them answering Alexa or Google any more, remote access gone. Belkin gave notice and offered refunds on units still under warranty. By the standards of this pattern that counts as the polite version.

The impolite version is Futurehome. The Norwegian company filed for bankruptcy on 20 May 2025, and the entity that picked up the platform put a mandatory annual fee on roughly 30,000 Smarthub II owners who had already bought their hardware outright. Refuse to pay and a firmware update strips out local network functions. Not a fee for a new service. A fee to keep the thing you own doing what it did last Tuesday.

And this is where I part company with the standard advice, which is to buy big brands because they will still be around. Belkin is a big brand, owned by Foxconn. Size did not save Wemo owners, because what killed those devices was not a solvency decision, it was a margin decision, and large companies make those faster than small ones. The same logic showed up when Microsoft locked Copilot to the account holder on a family plan: nothing broke, someone simply decided who was allowed through the door. Hardware now answers to the same reasoning, except the door is in your hallway.

The fee is the number worth sitting with. It buys nothing new: it is rent, applied after the sale, on a device whose purchase price already covered the hardware. Once one vendor proves a firmware push can convert a paid product into a subscription, trying it becomes a boardroom question rather than a legal one. Anyone comparing smart home gear on sticker price alone is reading half the invoice, roughly the mistake buyers made during the memory shortage that pushed device prices up.

Not every device dies the same way, and the differences are the whole game. Some Wemo units survived the shutdown untouched, purely because of how they were paired before the deadline. The table below separates what the vendor can revoke from what it cannot.

Read the right-hand column top to bottom and one line separates the survivors from the casualties: whether control ever had to leave the building. Everything routed through a company's servers died on schedule. Everything that spoke locally carried on without noticing.

Six months took the pattern from one bankrupt vendor to a household name, while the open standard that undercuts it finally added cameras and closures.

Where The Obvious Advice Falls Apart

"Just buy Matter devices" is the answer everyone reaches for, and it is directionally right and practically incomplete. Plenty of hardware carries a Matter badge and still routes the interesting parts, the automations, the history, the notifications, through the manufacturer's cloud. Matter guarantees you can switch the thing on and off locally. It does not guarantee the product you shopped for keeps existing.

There is a real grey area here that nobody has settled. When a company goes bankrupt, no one is obliged to keep servers running for hardware sold by a business that no longer exists, and someone has to pay for that infrastructure. But Futurehome's successor did not shut the lights off, it kept them on and billed for them, which is a different act entirely. Where the line sits between genuine cost recovery and a hostage situation is a question courts have barely started on. Worth remembering that Insteon simply vanished in 2022 with no notice, the version where nobody bills you and nobody helps you either.

Most of the practical failures are avoidable at the point of purchase rather than afterwards. Same discipline that separates a real backup from a hopeful one, covered in the case for moving off fragile drives: assume the convenient path will fail and check what remains. Cheap gear fails this test hardest, its own version of the slide in everyday product quality.

• Pair every device to a local platform on day one. Wemo owners who linked to Apple Home before the deadline kept control; the ones who waited got no second chance.
• Treat the vendor's own app as disposable. If a device only works properly through it, you are renting the hardware on a lease with no stated end date.
• Check whether automations run on a hub in your house or a server somewhere else. That detail alone decides what survives a shutdown notice.
• Keep the receipt. Refund eligibility during the Wemo wind-down turned entirely on warranty status, and most owners found out too late.

Go and open whatever smart home app you use, right now, and find out which of your devices still work if that app stops existing tonight. Anything that fails the test either gets paired to a local platform this week or gets treated as a purchase with an expiry date the seller has not told you yet. Buy the next one on that basis and the question stops mattering.

Related: what right to repair India actually covers when a device breaks