There is a line in The Last Samurai, spoken by Katsumoto about the cherry blossom: "The perfect blossom is a rare thing. You could spend your life looking for one, and it would not be a wasted life." The customer experience industry has spent the last twenty-five years searching for the perfect roll-up. On the evidence to date, the life has been wasted.

The Qualtrics acquisition of Press Ganey Forsta closes the longest M&A trail in the customer experience industry. It is also, on close inspection, the moment the industry quietly admits what most of its participants have spent twenty-five years failing to demonstrate: that the businesses underneath all this activity were largely unable to stand on their own two feet.

You can argue with that framing. I would invite you to try. The trail running from CustomerSat through MarketTools, Confirmit, Allegiance Technologies, MaritzCX, InMoment, Clarabridge, Market Metrix, Mindshare Technologies, Dapresy, FocusVision, Forsta, Press Ganey, and finally Qualtrics passes through an unusual number of private equity owners, an unusual number of integration narratives, and an unusual number of "transformations" that never quite converted into the software multiples the underwriting required. With two exceptions — Medallia and Qualtrics — almost everything in the customer experience industry has been a roll-up of companies that did not flourish independently.

I include my own former company, Satmetrix, in that verdict. We did real work, we built a real franchise, and we did not, at the end of it, stand alone in the way Qualtrics now stands alone. There is no embarrassment in saying so. The industry as a whole has the same shape, and most of us were participants in the same structural mistake.

The two outliers, and how they got there

Both Medallia and Qualtrics arrived at scale organically. The businesses underneath them, however, could not have looked more different — and the divergence is itself the most interesting story in the category.

Medallia: the diamond in the rough

Medallia did not look like a classic high-growth, high-margin SaaS business when it filed to go public. Total revenue of $313.6 million for the fiscal year ended January 2019, up from $261.2 million the year before — an annual growth rate of 20 percent. Roughly thirty percent of that revenue was professional services. The business had only just turned non-GAAP profitable, at a 6 percent operating margin in the most recent quarter. Net dollar retention was 119 percent — admirably sticky, but worth pausing on. Combined with overall growth of 20 percent, that retention rate implies almost all of the growth was coming from existing customers spending more, with very little contribution from new logos. That is a flattering picture of customer loyalty. It is a less flattering picture of new customer acquisition, and it raises a perfectly reasonable question about whether the growth rate was sustainable once existing-customer expansion eventually plateaued.

The model itself was not, exactly, a services business with software wrapped around it. It was the opposite: a software business with a high-touch service layer attached, the kind of service-enabled tech model that works very well for enterprise customers and makes most public-market investors skittish. It is far from the high-margin, high-growth, low-touch SaaS profile venture investors prefer. By the standard checklist VCs apply to enterprise software, Medallia is the kind of company you would expect them to pass on. And almost all of them would have.

Sequoia did not pass. Sequoia put roughly $105 million into Medallia across multiple rounds, owned around forty percent of the company at IPO, and saw that stake mark to roughly $1.8 billion on the first day of trading. By Sequoia's own framing this was one of the largest single bets in the firm's history. The reason it worked is the reason it should not, on paper, have worked. Sequoia recognized something the rest of the venture industry did not: that the high-touch, slower-burn financial profile was not a flaw of the business, it was the reason customers were buying. Medallia's customers were paying for an outcome that required a service wrap, and the wrap was part of the product. Sequoia bought a diamond in the rough and was rewarded for understanding what customers were actually willing to pay for. That is, increasingly, a rarer skill than the venture industry likes to admit.

Qualtrics: a misfiled enterprise software company

Qualtrics looked like the opposite. The same vintage of public filings tells a different story. Revenue growing at around 40 percent year-on-year going into the SAP deal. Subscription gross margins in the high eighties. Overall blended gross margins in the mid-seventies. Professional services running at roughly a quarter of revenue rather than a third. Free cash flow positive, with margins in the mid-teens. A Rule of 40 score in the region of 50. By every metric on the SaaS dashboard, this was a textbook software company, and the public markets were, by 2018, plainly ready to value it as one. Where Medallia ran a software business with a heavy service layer optimized for big enterprise accounts, Qualtrics ran a software business with almost no service layer optimized for individuals and small teams. Same category, opposite operating model.

The interesting thing about Qualtrics is that it was not, in any meaningful sense, an enterprise software company at all. The actual sales motion was lot-size one — single seats sold to individual researchers, marketing analysts, and academic departments who needed to run a study. In my view the closer historical analogue is not Salesforce. It is Adobe. Adobe grew up selling single creative seats to individuals, then patiently learned how to translate that motion into enterprise sales, and proved it understood the crossover when it bought Marketo in 2018. My view, for what it's worth, is that Adobe would have been the natural home for Qualtrics. The deal that happened was not that one.

SAP arrived in the eleventh hour with $8 billion to keep Qualtrics out of the public markets entirely. Acquiring a company days before its planned IPO, at a multiple the public market had not yet had the opportunity to disagree with, is the kind of move that can charitably be described as enthusiastic. The transaction has to count as one of the most blindingly unobvious purchases in modern enterprise software. By any reasonable reading Bill McDermott was steamrollered into it, and he left SAP within a year of closing it.

The round trip is more interesting than it first appears. SAP put $8 billion in. It took roughly $1.55 billion out in the January 2021 spin-off IPO at $30 a share, and a further $7.7 billion at $18.15 a share when Silver Lake took the company private in 2023 — a total of around $9.25 billion against the $8 billion cost base. That is a nominal gain of about 16 percent over four-and-a-half years, or roughly 3 percent annualized — broadly what SAP could have earned by leaving the $8 billion in a money-market fund and taking up gardening. Adjust for the cumulative dollar inflation of the period and the cost of capital SAP could have earned holding almost anything else, and the real return was negative.

The more revealing part is how SAP got to a near-break-even at all. Qualtrics in 2018 was bought at roughly twenty times revenue, the kind of multiple only available in a SaaS-buoyant market and only paid by a strategic acquirer trying to head off an IPO. The 2021 spin-off landed in the dead center of the post-COVID software bubble and got a slice of the position out at something like thirty times trailing revenue — timing-lucky in the simplest sense. By 2023 the multiple had collapsed, the strategic-buyer premium was gone, and SAP sold the rest at about nine times revenue into a leveraged take-private. Operationally, Qualtrics had tripled its revenue under SAP's ownership — from roughly $400 million at deal close to something north of $1.4 billion by 2022 — and the reason none of that operational growth showed up in SAP's return is that the multiple compressed faster than the revenue could grow. SAP was bailed out by the 2021 bubble for the first slice and punished by the 2023 trough for the rest, and was saved from a much worse outcome by the operational work Ryan Smith's team produced inside the SAP umbrella. Which, uncharitably, is another way of saying the value SAP got out of Qualtrics was largely the value Qualtrics created for itself.

The public-market shareholders who bought the 2021 spin-off received the other side of that arithmetic. The IPO priced at $30 and opened at $41.85. Anyone who bought in the offering and held through to the Silver Lake take-out received $18.15 — a forty percent loss for the IPO buyer and closer to fifty-seven percent for anyone who chased the first-day pop. The seventy-three percent premium Silver Lake paid over the unaffected price tells you precisely how far the stock had to fall before a buyer was willing to step in. The bail-out worked. The public shareholders paid for it.

Qualtrics never found a home at SAP. The total value creation across the round trip, after inflation, opportunity cost, and the management distraction the deal carried, is not a number SAP would put on the cover of an annual report.

Ryan Smith, on the other hand, deserves the kind of respect founders rarely get for an outcome like this. He took care of his shareholders, and himself, with surgical timing. The fact that the deal made very little sense for the buyer is not his problem. He may not always get the credit he is owed for what he pulled off here. He certainly got the payoff. One suspects he has made his peace with the trade-off.

What Qualtrics actually won

Qualtrics's real victory is something almost nobody, at the time of the SAP deal, would have called the win. It is a commoditization play. Qualtrics, more than anyone else in this category, turned survey software into a product that essentially sold itself — distribution-led, into the long mid-market tail nobody else in the industry could profitably service. It won by becoming the cheap, easy, good-enough version of what the rest of the industry was charging six and seven figures for. That is the same playbook Adobe ran in creative tools. It is the playbook the rest of the industry refused to acknowledge was the actual game being played, while it was busy buying itself.

If you wanted a one-sentence verdict on the last twenty-five years it would be this: in a category that was always going to commoditize, the winner was the company that committed to commoditizing it.

What Medallia just found out

Medallia, in the meantime, has just found out something rather more decisive than a difficult patch. Thoma Bravo took the company private in 2021 for $6.4 billion, financed with the kind of leverage that only makes sense in a low-interest-rate world. The interest-rate world that followed was not that one. Growth slowed under the new ownership, interest expense climbed alongside the rate cycle, and concerns about AI disruption to the underlying methodology compounded the financial pressure. Last week, Thoma Bravo agreed to hand Medallia over to its lenders — Blackstone, KKR, Apollo, and Antares — in a debt-for-equity restructuring that wipes out roughly $5.1 billion of equity value held by Thoma Bravo and its co-investors. The company's $3 billion of debt becomes its new equity, and a credit syndicate becomes its new owner. Private equity firms occasionally part ways with their portfolio companies. They do not, as a rule, do so by handing the keys directly to the people they borrowed from.

The cause is not subtle. The reason Medallia customers liked Medallia was the service layer that a financial model would naturally treat as overhead. A turkey, in the week before Thanksgiving, looks like an obvious opportunity to a private equity owner. The problem is that this particular turkey was the goose, and the eggs were the relationships. Thoma Bravo took the goose. It is now finding out, with very public consequences, what was actually left behind.

Whether Medallia, now in the hands of its credit-fund owners, can adapt to a commoditizing market that increasingly looks like a standard set of software products with low-touch services is a question for the next owner. What is clear is that the Qualtrics model has, for now, been the winning one — if not in the enterprise, then in the mass market the enterprise has to compete against eventually.

The long tree of weak roots

Everyone else has, to varying degrees, been propped up. The Confirmit tree, the InMoment tree, the Maritz tree, the Allegiance tree, the Press Ganey tree — each followed the same shape. Take a company struggling to break out of its niche. Combine it with another company struggling to break out of its niche. Tell a private equity owner the combined entity now has a "platform" story. Sell it to the next private equity owner with a slightly bigger platform story. Repeat until either the music stops or a strategic acquirer turns up to write the final check.

The aggregate value created by all of this activity, from the perspective of the businesses themselves, is genuinely hard to identify. Working it out properly would take a forensic effort no one is going to fund. Each individual transaction probably made sense to the parties at the time. Each multiple expansion probably looked plausible against the comparable. But step back and look at what was actually built across twenty-five years and the picture is striking: very little independent strength, a great deal of integration cost, and a long tail of customers who were inherited rather than won.

Where the value actually lived

The most uncomfortable observation, for an industry that spent a quarter of a century talking about platforms, is that the durable value almost never sat in the platform. It sat in the data.

Press Ganey's franchise — thirty-five years of patient experience benchmarks across forty-one thousand healthcare facilities, fused to an HCAHPS regulatory mandate hospitals could not work around — is the actual asset. The survey software wrapping it is largely incidental. MaritzCX's franchise was, similarly, the New Vehicle Customer Study, run continuously since 1969, used by ninety percent of North American OEMs because the methodology had become the methodology.

When Qualtrics writes a $6.75 billion check to acquire Press Ganey Forsta, it is almost entirely a check for the healthcare benchmark dataset and the regulatory moat that protects it. The Forsta survey technology — the thing the roll-up was meant to assemble — is either redundant against what Qualtrics already has, or considered inferior to it. Twenty-five years of platform integration, in other words, has been valued by the buyer at approximately the cost of removing it. The industry spent twenty-five years optimizing for software multiples and building software companies, and at the exit, the strategic buyer turned up and paid for the data.

The investor question, briefly

You can ask, as I occasionally do, whether the financial engineering layered on top of all this produced returns commensurate with the risk and effort. An earlier draft of this article tried to answer that question forensically — tracing each entity through each owner, normalizing the multiples, adjusting for inflation and the cost of capital, and arriving at a defensible view on whether the activity in aggregate created or destroyed shareholder value. The exercise was abandoned. The math turned out to be both intractable and not, in the end, worth anyone's time.

That, in itself, is informative. When the financial outcome of an industry is genuinely that hard to compute, the result is almost certainly a poor one. The financial-services profession is exceptionally well-equipped to publicize wins and exceptionally creative at burying losses beneath several layers of legitimate-sounding analytical complexity. The fact that nobody has been publishing celebratory papers on the spectacular shareholder returns generated by CX roll-up strategies tells you most of what you need to know about how those returns actually went. The probable answer is that they went rather worse than most of the participants would care to acknowledge.

What can be said with reasonable confidence is that the larger PE trades — EQT's Press Ganey hold, the subsequent Ares and Leonard Green hold — produced adequate but not exceptional outcomes, often driven more by multiple expansion than by anything that happened operationally inside the businesses. The smaller trades, on undisclosed terms, were probably modest. Thoma Bravo's Medallia trade is now, very publicly, a $5.1 billion equity wipeout. The aggregate pool of management attention, integration labor, debt service, and customer disruption this M&A activity consumed, set against the value created, would not flatter the industry if it were ever properly measured. In the technical language of the trade, this is sometimes referred to as a learning experience.

The opportunity cost no one talks about

Which brings me to the secondary thesis, and the one that should worry anyone still operating inside this category.

While the industry was busy integrating, it was not busy innovating. Twenty-five years of roll-up activity is twenty-five years of management attention spent on platform consolidation, customer overlap analysis, system migration, and integration roadmaps. It is not twenty-five years of rethinking from first principles what customer measurement should actually do. The underlying technology of the industry — survey design, distribution, response collection, dashboarding, NPS scoring — looks remarkably similar today to how it looked in 2003. The interfaces are prettier. The cloud bill is bigger. The methodology is the same.

You could think of it as a generation of incumbents who, at exactly the moment AI made it possible to rebuild the entire category from first principles, were still drawing org charts and cleaning up post-merger CRM overlap. Disruption tends to find industries in this state. The Medallia handover to its creditors last week, the Qualtrics financing wobble at JPMorgan, the public market's increasingly impatient pricing of survey-based experience management as a commoditizing category — these are not anomalies. They are the bond market and the equity market beginning, slowly, to price the same thing.

The verdict

The data points one direction. The only models that produced durable value in this category were the two organic businesses that committed to a clear thesis about what customers would actually pay for, and then built the company that thesis required. Sequoia recognized that Medallia's customers were paying for outcomes that needed a service wrap, and built the business around the wrap. Qualtrics committed to volume and commoditization, distributed itself into the long mid-market tail, and won by becoming the cheap, easy version of what the rest of the industry had been overcharging for. Neither of these is a financial-engineering outcome. Both are the result of selecting a thesis, executing it for fifteen-plus years, and letting the customer relationship compound.

Everything else in the tree was an attempt to manufacture, through M&A, the appearance of a platform business on top of underlying companies that were never built to carry one. The data is consistent with a story in which roll-up activity preserved data assets that already existed, generated mediocre returns largely from multiple expansion, and produced almost nothing in the way of new operational capability. Press Ganey's exit value at $6.75 billion is overwhelmingly a payment for thirty-five years of healthcare benchmark data that accumulated regardless of who owned the platform around it. The platform was the part that did not earn its keep.

The position this leaves the industry in is uncomfortable. The two organic winners are now visibly carrying their own version of trouble. Medallia has just been handed to its lenders in a $5.1 billion equity wipeout — the most public failure of a private-equity owner in the history of this category. Qualtrics is now carrying $6.75 billion of healthcare-data goodwill at a moment when the bond market is openly pricing AI disruption to survey-based experience management — its own debt deal at JPMorgan was halted in front of investors who decided they did not want to underwrite the assumption. The roll-up tree carries the heaviest integration debt and the oldest methodology in the category. None of these is the position you would want to occupy if a credible AI-native alternative arrived.

The conclusion the data supports is the unflattering one. Twenty-five years of activity in this industry produced two businesses worth keeping, a quantity of benchmark data that survived its successive owners, and a structural setup that is now plainly ripe for whatever comes next. If the next thing arrives, it will not arrive from inside the tree. It will arrive from outside, built on a different substrate, by people who do not have the burden of integrating anyone else's history.

The moral, briefly

Across twenty-five years of activity in this category, who actually made money is a more compact story than the M&A history would suggest. It can be a wonderful business being a venture investor, if you can persuade the public market to absorb your risk at a generous premium. It is a miserable business being a public investor if you happen to be on the other side of that trade. Being a corporate acquirer in this category has been, throughout, a caveat emptor situation in which enthusiasm has reliably trumped logic. Investment bankers, who collected fees on the way in, fees on the way out, fees on the spin-offs, fees on the take-privates, and fees on the restructurings, have done very well at every step, regardless of which direction the assets were moving. As for the employees who built and ran the businesses through all of it — on reflection, they were probably better off collecting cherry blossoms.