Go back on a position you staked out in public, or keep defending it while the facts move.
11 of 15 judged reversals stuck. 4 were themselves undone, one of them within 79 days.
The reversals that stuck named what had changed underneath. The cost of going back was set by what the first move had bought.
The promise Netflix put in writing
In 2019 Netflix told its shareholders, in writing, that it would not sell advertising: "We, like HBO, are advertising free." Anyone who heard otherwise, the letter said, could be sure it was false.
In April 2022 Netflix reported that it had lost subscribers. At the earnings interview that same day, Reed Hastings floated the idea of ads. In July Netflix named Microsoft its advertising partner. On November 3, 2022 it launched "Basic with Ads" at $6.99 a month, $3 less than the ad-free plan.
Four weeks later Hastings explained the turn in one sentence.
“I didn't believe in the ad-supported tactic for us. I was wrong about that.”Reed Hastings, November 2022, as recorded in the Netflix case
Over the next three years Netflix's shares beat the S&P 500 by 182.9 points. Not every reversal ends that way, and the share price is not the point of the story. The point is what Netflix did: it named what had changed, said plainly that it had been wrong, and moved fast.
That is the decision in one story. A company stakes out a position in public. The facts under it move. Now it has to choose between its word and its business. Plenty of companies in this guide reached the same choice, and 4 of them ended up taking their reversal back. The rest of this guide is about why some changes of mind stick and others don't.
What strategic reversal is
A strategic reversal is when a company publicly abandons a position it had staked out, such as a promise, a product or a policy, and does the opposite.
It is different from an ordinary change of plan because people outside the company were relying on the old position: customers who chose you for it, investors who priced it in, partners who built on it. So a reversal has two costs. One is in the business. The other is in the word you broke, and it is paid by whoever has to stand up and say so.
Not everything called a reversal is one. Some moves only look like a U-turn from the headline, and reading what was actually promised is the first step in judging one.
Why it is so hard
On paper, changing course when the facts change is plain good sense. In a real company, three things make it slow.
You said it in public
In June 2022 Atlassian wrote: "To be clear, we have no plans to end of life or end of support for our Data Center offering." In September 2025 it set an end-of-life date for Data Center. Every public promise is a sentence someone will quote back to you.
The person who has to reverse is the one who said it
Admitting the mistake is personal. When Reed Hastings explained why Netflix had held out against ads, he named his own belief: "I bought in for a decade to the belief that systems relying on data were going to be able to do higher CPMs than anyone else." It took a subscriber loss, and some convincing from his finance chief, to move him.
The first numbers look like failure
When Adobe stopped selling boxed software in 2013, its operating income fell from $1.18 billion to $423 million in a single year. A reversal that is working and one that is failing look the same for the first few quarters.
The ways companies did it
The reversals on record take four shapes. They differ in how openly the company admits the change, and in who forces it.
Say what changed, then switch
Name the fact that moved, and do the opposite of what you said.
- Microsoft
- Had fought Linux to protect its license revenue. Once it made money from customers running anything on its cloud, it switched sides, and in 2018 paid $7.5 billion in stock for GitHub. By April 2024 more than 60% of customer cores in Azure ran Linux.
- Netflix
- Tolerated password sharing for most of its life, then began enforcing against it in the U.S. on May 23, 2023. Daily sign-ups reached 73,000, a 102% jump, and it added roughly 50 million paying subscribers in the eighteen months that followed.
Remove the old option on a date
Stop selling or supporting the old thing, so customers have to move.
- Adobe
- Announced in May 2013 that there would be no more perpetual licenses, only Creative Cloud subscriptions. Product revenue fell from $3.34 billion to $2.47 billion in a year; in 2014 subscription revenue of $2.08 billion overtook it.
- SAP
- Set a hard end to support for its old ERP software, with paid extended maintenance only through December 2030, after moving its deadlines several times before. By the end of 2024 about 14,000 of 35,000 customers had migrated.
Pull back under pressure
Drop a public commitment, or a market you fought for, once it stops paying.
- BP
- With Elliott Management holding a ~5% stake, scrapped its 50 GW renewables target in February 2025, cut transition spending to $1.5–2 billion a year and lifted oil and gas investment toward $10 billion a year.
- Southwest
- Its September 2024 Investor Day kept bags flying free. After a settlement added six Elliott-backed directors, it announced bag fees in March 2025: $35 for the first checked bag, the first bag charge in its 54-year history.
- Uber
- Had fought for Southeast Asia against Grab, a local rival that launched in 2012 with 40 drivers in Malaysia. On March 25, 2018 it bowed out, selling its operation in several countries to Grab.
The reversal that wasn't
A headline calls it a U-turn; the fine print says otherwise.
- Apple
- Launched the $99 Apple Pencil in 2015 and was mocked with Steve Jobs's line "If you see a stylus, you blew it." But the Pencil was sold separately and never required, and Jobs had been talking about phones.
- Volkswagen
- Presented its €180 billion plan for 2023 to 2027 as an electric pivot, while reserving about a third, roughly €60 billion, for combustion engines. Its finance chief said in 2024: "It is a third and it will stay a third".
What 19 companies did
Here is every decision on record, placed by the year it was made and marked by how it turned out. Hover a mark, or open the table, for the company and the outcome.
19 reversing course decisions, 1985 to 2025
Each mark is one decision. "Never undone" means the company did not reverse it, not that it worked.
Show as a table
| Company | Year | Choice | Outcome |
|---|---|---|---|
| Coca-Cola | 1985 | Reformulate to sweeter recipe | Walked back |
| Adobe | 2013 | Go subscription-only | Never undone |
| Apple | 2015 | Ship optional add-on stylus | Never undone |
| AT&T | 2016 | Buy content and integrate | Walked back |
| Microsoft | 2018 | Embrace and host open source | Never undone |
| Uber | 2018 | Exit and sell to Grab | Too early to call |
| SAP | 2020 | Set hard end-of-support date | Never undone |
| Hyundai | 2022 | Build dedicated US EV plant | Never undone |
| Netflix | 2022 | Launch an ad tier | Never undone |
| Walmart | 2022 | Open the seller floodgates | Never undone |
| Airbnb | 2023 | Freeze new submissions only | Walked back |
| Netflix | 2023 | Enforce a crackdown | Never undone |
| Salesforce | 2023 | Cut 10% of staff | Never undone |
| Volkswagen | 2023 | Dual-track hedge | Never undone |
| Shell | 2024 | Prune enforceable targets, keep net-zero | Too early to call |
| Universal Music | 2024 | Pull catalog off TikTok | Never undone |
| Atlassian | 2025 | End-of-life Data Center | Too early to call |
| BP | 2025 | Retreat to oil and gas | Too early to call |
| Southwest | 2025 | Charge bag fees | Walked back |
Of the 15 decisions that can be judged, 11 stuck: the company never went back to the old position. 4 were themselves undone, and 4 are too recent to judge.
Sticking is not the same as working. Walmart's opening of its marketplace stuck, and by mid-2025 it had crossed 200,000 active sellers; a CNBC investigation that September found it had made vetting "more lax in a bid to compete with Amazon," and uncovered counterfeit products on the platform.
The walk-backs
Three reversals on record were themselves taken back. Put them side by side and the gap between them is not how wrong each company was. It is how long it took to go back, and what had to be sold to do it.
| Company | What it did | How it was undone | Kind of move |
|---|---|---|---|
| Coca-Cola1985 | Replaced its 99-year-old formula with a sweeter one that had won blind taste tests | Brought the original back as Coca-Cola Classic 79 days after launchHotline calls of some 8,000 a day by June, and roughly 40,000 letters | Undone by customers |
| AT&T2016 | Agreed to buy Time Warner, after DirecTV, to own the content as well as the pipes | Let WarnerMedia go in 2022 for $40.4 billion in cash and retained debt; sold the last 70% of DirecTV in 2025Net debt around $156 billion at the end of 2021 | Undone by the balance sheet |
| Airbnb2023 | Froze new Experiences submissions to protect the core business | Reopened submissions after a roughly 17-month pause; relaunched Experiences in 2025 with a new Services layer in 260 citiesRoughly 5,000 listings removed, with their forward reservations canceled | Paused by design |
Southwest's 2025 bag fees are marked as reversed in the record because they broke a public pledge; the fees themselves were never taken back, so they are not shown here. The kind of move is Stratrix's reading of each case.
Coca-Cola changed a recipe, and a recipe can be changed back. Airbnb froze an intake valve rather than closing the business, so reopening it was a decision, not a retreat. AT&T had committed more than $175 billion in announced deal values, $67.1 billion for DirecTV and $108.7 billion for Time Warner, and getting out meant selling businesses one at a time until 2025.
The cost of going back was set by what the first move had bought. A move that buys nothing can be reversed in weeks. A move that buys companies is reversed by selling them.
A recipe can be changed back in 79 days. A $108.7 billion acquisition can't.
What the share price did
For many of the decisions there is a clean market record: the company's total return over the three years after the decision, against the S&P 500.
Three years after the decision, against the S&P 500
Percentage points of total return above or below the index, for the 8 decisions with a clean market record. 88% beat it; the median was +62.5 points.
Show as a table
| Company | Year | vs S&P 500, 3 years | Outcome |
|---|---|---|---|
| Netflix | 2022 | +182.9 | Never undone |
| Microsoft | 2018 | +91.4 | Never undone |
| Walmart | 2022 | +75.8 | Never undone |
| Adobe | 2013 | +71.7 | Never undone |
| Apple | 2015 | +53.2 | Never undone |
| Netflix | 2023 | +50.6 | Never undone |
| Salesforce | 2023 | +7.3 | Never undone |
| Airbnb | 2023 | −66.0 | Walked back |
Read the chart as a picture of the companies, not a score for the reversals. Salesforce's 2023 job cuts stuck, and its non-GAAP operating margin reached 30.5% in fiscal 2024, past the target it had set; over three years its shares beat the index by only 7.3 points. A reversal can do what it promised and the stock can still barely move.
How to run this decision
If you are weighing a version of this choice, here is what the record suggests doing, in order.
- Write down what changed underneath, not that you changed your mind
A reversal people can follow names the fact that moved. Microsoft's money moved to the cloud, where a Linux customer pays rent. Netflix lost subscribers. If you cannot name the fact, you are not reversing a strategy, you are losing an argument.
Seen in: Microsoft, 2018 · Netflix, 2022 - Check your own published numbers before you contradict them
Southwest's 2024 Investor Day slides put bag fee revenue at most $1B to $1.5B a year and the share it would lose at ~$1.8B a year. Five months later it charged for bags. If your own figures argued for the old position, say what changed in the figures.
Seen in: Southwest, 2025 - Say it once, plainly, in the first person
Hastings said he had been wrong, gave the reason, and named who had persuaded him. That ends the argument about whether the company has really changed course, which a vague statement keeps alive for years.
Seen in: Netflix, 2022 - Price the trough before you enter it
Adobe's operating income fell by most of its size in one year before subscriptions overtook boxed sales. Write down how deep the dip should be and when it should end, so a bad quarter reads as the plan working rather than the plan failing.
Seen in: Adobe, 2013 - Use a deadline when persuasion stalls, and plan for the holdouts
A date moves customers that arguments don't. But SAP moved its own deadline several times, and by the end of 2024 most of its old customers had still not migrated. Decide in advance what you will do about the ones who never move.
Seen in: SAP, 2020 · Atlassian, 2025 - Take a pause instead of an exit when you are unsure
Airbnb froze new Experiences rather than shutting them, pruned the weak listings, and reopened about seventeen months later. A pause keeps the option to come back; an exit sells it. Say at the start what would reopen it.
Seen in: Airbnb, 2023 - Before the first move, write down what it would take to undo it
Coca-Cola could bring its old formula back in weeks. AT&T had to sell WarnerMedia and DirecTV. The cost of a future reversal is part of the size of today's decision, and it is rarely in the paper the board approves.
Seen in: Coca-Cola, 1985 · AT&T, 2016
Warning signs
What it has looked like when this decision was going wrong.
- Your own investor presentation argues for the position you are about to drop.Southwest, 2025
- A product line carries a public promise of the "we have no plans to end" kind.Atlassian, 2022 and 2025
- The change of course arrives with an activist, not with a change in the business.BP, 2025 · Southwest, 2025
- The announcement is bigger than the money behind it.Volkswagen, 2023
- Undoing the first move would mean selling a company.AT&T, 2016
All 19 cases
Each case is one company at one decision: what was on the table, what it chose, why, and what followed, with every fact sourced.
Never undone · 11
- Adobe, 2013Go subscription-only
- Apple, 2015Ship optional add-on stylus
- Microsoft, 2018Embrace and host open source
- SAP, 2020Set hard end-of-support date
- Hyundai, 2022Build dedicated US EV plant
- Netflix, 2022Launch an ad tier
- Walmart, 2022Open the seller floodgates
- Netflix, 2023Enforce a crackdown
- Salesforce, 2023Cut 10% of staff
- Volkswagen, 2023Dual-track hedge
- Universal Music, 2024Pull catalog off TikTok
Walked back · 4
- Coca-Cola, 1985Reformulate to sweeter recipe
- AT&T, 2016Buy content and integrate
- Airbnb, 2023Freeze new submissions only
- Southwest, 2025Charge bag fees
Too early to call · 4
- Uber, 2018Exit and sell to Grab
- Shell, 2024Prune enforceable targets, keep net-zero
- Atlassian, 2025End-of-life Data Center
- BP, 2025Retreat to oil and gas
Questions
- What is a strategic reversal?
- When a company publicly abandons a position it had staked out, such as a promise, a product or a policy, and does the opposite. Netflix launching an ad tier after years of saying it would stay ad-free is a clear example.
- Is reversing a strategy a sign of failure?
- Not by itself. A reversal that names what changed and moves quickly can be the best decision a company makes. The expensive reversals are the ones that come late, after the company has bought assets it then has to sell.
- What are examples of companies that reversed course?
- Netflix on advertising and on password sharing, Microsoft on open source, Adobe on boxed software, Coca-Cola with New Coke, and Southwest on free checked bags. Coca-Cola's New Coke is the clearest case on record of a reversal that was itself undone, within 79 days.
- How do you reverse a strategy without losing trust?
- Name the fact that changed, say plainly that the old position was wrong, and act quickly. Where you are unsure, pause rather than exit, and set out in advance what would make you reopen.
More from Stratrix
“When do you reverse a position, and what does a year of waiting cost?”
See the pack — $499 →All Access
Every Casebook, three Strategy Packs a seat each year and the Analyst, for teams that face these decisions all year.
$1,999 a year · $499 a seatWhat is included →Weekly email
Stratonomics, every Tuesday
One data story every Tuesday: what happened, what it means, and the figures you can quote. Free.