From ¥1 Billion to ¥50 Billion: A Medical Data Company Needed Ten Years to Reach the Second Half

A plan for one year — plant grain. A plan for ten years — plant trees. A plan for a lifetime — cultivate people.

—— Guanzi, “Quanxiu” chapter (Warring States period; translated by the author)

What This Episode Is About

The 16 August 2026 episode of Investor’s Sunday features Yosuke Matsushima, chairman of JMDC. The conversation follows his career: he started at Dai-ichi Life, spent years at two large consulting firms, moved into private equity, and in 2012 joined a photo-processing equipment maker as vice president with a mandate to turn it around. From that seat, in 2013, he bought a small company called Japan Medical Data Center — revenue around ¥1 billion, profit hovering near ¥100 million.

Twelve years later, that company’s revenue is ¥50 billion.

The back half of the episode is just as interesting. Having stepped up from CEO to chairman, he started a very small investment firm with his former right hand, funded entirely out of his own pocket. Five investments so far — one of which had EBITDA of negative ¥600 million when he took it over.

The thing I wanted to write about isn’t “what great instincts.” He says plainly on the show that at the moment of purchase, ¥50 billion was unimaginable to him. The question worth unpacking is different: when you can’t see the destination, what actually carries you through twelve years?

Key Points

1. He bought at the moment someone else was forced to sell — in a shape nobody else could offer.

The seller was Olympus. After the loss-concealment scandal, the company was placed under supervision and had to clean up its investment portfolio before a deadline. The large assets sold first, leaving a pile of small ones nobody wanted. Matsushima’s proposal: I’ll take all ten together. That proposal was only possible because he was deploying the parent company’s own money, not fund money. A normal fund has to justify every investment to its LPs; “and nine others I don’t especially want” is not a sentence you can say.

2. A data business has a hooked cost curve, and he bought on the day the hook turned up.

The company burned ¥3 billion over its first decade and had only just started turning a profit when he acquired it. Matsushima explains it cleanly: before you reach scale, data is brutally expensive — “we have data on ten thousand people” is not a sellable sentence. But past the threshold, reproduction and transfer costs are essentially zero, and margins above fixed costs become extraordinary. The painful first decade and the sweet second one are two ends of the same curve.

3. So his first move was pushing the dataset from 500,000 people to 5 million.

Not because bigger is nicer, but because of a harsh property of data businesses: the value lives in the largest dataset. Second place isn’t half the value; it’s a rounding error. Scaling did two jobs at once — raised what he could charge, and put the field out of reach. That isn’t marketing language; it’s the physics of the business.

4. The most counterintuitive part: he doesn’t buy data. The data providers pay him.

Japan’s health insurance societies traditionally handle membership records and payment processing — administrative work. As national healthcare finances tightened, they were told to use data to manage members’ health and promote generic drugs. They had no idea how. Deciding which of tens of thousands of members to say what to is genuinely hard. JMDC took the job on: calculating exactly how much switching to generics would save you, flagging which health-check number has deteriorated and what to watch. Priced deliberately cheap, in exchange for the right to anonymize and commercialize the data. The supply side became the revenue side.

5. The real ceiling isn’t the data. It’s literacy.

Matsushima puts it bluntly: hand people data and most of them don’t know how to look at it. So he built a consulting arm to translate. For pharmaceutical clients, they map “patient journeys” — before a diagnosis is finally made, what symptoms typically appear first, which tests get ordered, what comorbidities tend to follow. A sales rep armed with that map can explain to a physician why a given drug belongs at a given point. For rare diseases, individual case counts are tiny, but aggregate data can establish roughly how many people nationwide are affected — and only then does an R&D budget get approved. Same data, one layer of translation, entirely different price.

6. He deliberately keeps big data away from the bedside.

This was the most disciplined passage in the episode. In clinical practice there’s one patient in front of you, and even a 10% probability doesn’t let you say “socially this doesn’t pencil out.” So his judgment is that big data belongs first in insurance systems and hospital operations — places where volume creates economics — and not in clinical decisions. Clinical accumulation becomes big data; the reverse doesn’t follow. Knowing where your tool stops working is harder than knowing where it works.

7. The shape of 200x: 4x in the first five years, 50x in the next five or six.

This is the number I’d most want remembered. From 2013 to 2018, five years, four times. Matsushima’s own comment: “The important part is that first 4x — an ordinary private equity fund would have exited right there.” The real move came afterward. His explanation: give a good company twice as much time and the strategic options available to it keep multiplying.

The operating discipline that goes with it: get EBITDA margin to 20–30%, and above that, let me plough it back into the next round. He treated growth rate plus margin as his own report card, aiming for 30 plus 30 equals 60 — because growing furiously with no profit and printing profit with no growth are both, to him, out of balance.

Going Further

”The fundamentals improved but the stock didn’t move” — you may have bought the first half of the curve

This is a real situation for a lot of people. After you buy, the company genuinely improves, the numbers get better quarter after quarter, and the market does nothing. Eighteen months in, you start doubting yourself.

This episode offers a concrete reference point. Before the acquisition, JMDC had burned a decade and ¥3 billion to barely squeeze out a profit. Standing at year eight, what you’d see is a small, chronically loss-making company with ¥1 billion in revenue. Only by opening up the cost structure would you learn that the decade wasn’t failure — it was a threshold. The distinguishing question is: past fixed costs, how much profit does the next dollar of revenue carry? That’s visible in the financials, but you have to go looking. It isn’t written in the growth-rate column.

So here’s the question in a form you can use: am I waiting on something time will fix, or something time will worsen? The dividing line isn’t emotional; it’s whether a mechanism exists that accumulates chips for you while you wait. Every additional insurance society JMDC served thickened its dataset, and the dataset was the moat — waiting was doing work on its behalf. Conversely, if the trouble is eroding demand, closing competitors, or compressing prices, then waiting is doing work for the other side.

One level deeper: this same test tells you when to admit you were wrong. Not at a percentage decline, but when the mechanism that made waiting worthwhile breaks. In this case, that mechanism is the lead in dataset size. The day that gap starts converging, all the prior patience loses its basis — and that’s the signal to recalculate, regardless of what your position shows.

”The news calls it a data play / an AI play” — this episode hands you three tests

The single highest-value sentence in the interview: plenty of companies hold all sorts of data, but companies that turn data into a business barely exist. When he agreed to join another company’s board as an outside director, what he was looking at was the capability to convert big data into a business — not the possession of big data.

Read the episode backwards and you get three questions to run on any self-described data-driven company:

One: does this company pay for its data, or get paid to collect it? JMDC is the latter. That determines more than cost — it determines renewal stickiness, because what the customer is buying is “do a job I can’t do,” with data as the byproduct. A company that must purchase its data has no structural gap versus competitors; whoever bids highest gets it.

Two: is the scale overwhelming? There’s no runner-up premium in data. So “20% share, top three in the industry” — fine in other industries — deserves a question mark here.

Three: have customers reached the point where they don’t dare decide without it? Matsushima described that progression: the world ran fine without it; then people tried it and thought “fair enough”; then it became impossible to decide correctly without looking. That’s observable. Is the product decoration on a report, or a step nobody can remove from the workflow? Getting there depends on the consulting arm, not the database.

Few companies pass all three — which is the point. It turns “data play” from a slogan back into a checkable question.

”Does my money have an expiry date?” — this decides the outcome before stock selection does

The closing section stayed with me. Matsushima runs his new investment firm entirely on his own capital, and his reason is: take outside money and you have to return it in three to five years. “But that’s not what we spent all those years doing.”

Put that next to the 200x and it clicks. Four times in five years — a fund exiting there is entirely reasonable. Not bad judgment; the fund’s life forced the timing. Same company, same thesis, and only because the money had a different expiry date, the outcome differed by fifty times.

For an individual investor this lands more directly than it first seems. The real problem usually isn’t reading the company wrong — it’s using money with a deadline to buy something that only pays if you have no deadline. That money is a down payment in six months, tuition next year, or just “I want a result before year-end.” All deadlines. And deadlines disguise themselves as judgment: you believe you’re cutting a loss rationally, when in fact you simply couldn’t last until the date.

So the order of questions before you buy is the reverse of the usual one. Not “is this company worth three years of waiting,” but “can I leave this money alone for three years.” Get the second one wrong and getting the first one right doesn’t help.

Further Reading

• The episode itself: Investor’s Sunday, 16 August 2026, with JMDC chairman Yosuke Matsushima. The previous week’s episode covers the first half of his story; the two work better together.
• JMDC’s quarterly earnings presentation decks. Matsushima mentions writing them himself, quarter by quarter, putting the period’s problems and responses into words — an operator-written deck deliberately made legible to employees is good reading material in its own right.
• Japan’s health insurance society system and the concept of the PHR (personal health record) — necessary background for why this business collects from both sides.
• Japan’s business turnaround ADR framework, which he used when personally sponsoring an advertising and photography production company. He notes that 120 creditors attended the meeting, and the sponsor’s chair held exactly one person.

One Thing to Take With You

One idea: outcomes are decided less by whether your judgment is right than by whether your resources have an expiry date.

Same call, same company: money due in three years exits at 4x; money with no due date reaches 200x. The gap isn’t intelligence, it’s structure. This holds well outside investing. A relationship, a skill, a piece of work nobody has noticed yet — the moment most people quit is the moment the deadline they privately set for themselves arrives, not the moment the thing actually failed.

One exercise you can do today: write an honest expiry date for three things you’re currently waiting on.

Not “someday.” The real date — the one you set in your head, after which you’ll quietly think forget it. It might be waiting for a position to break even, waiting for training to show results, waiting for a strained relationship to soften, waiting for something you made to be seen. Three things, three dates, on one piece of paper.

Then pick one and double the date. Ask yourself: if I had that much time, what would I do differently today?

Most people find the answer is concrete, and usually some version of “I wouldn’t be forcing a result this fast” — the thing that had been compressed into a sprint grows back into the shape it should have had. The real output of this exercise isn’t the new date. It’s finally seeing that the urgency was self-issued.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.