Company history

Sansan was registered in Tokyo in June 2007 with one stated purpose: business card management. Its founder, Terada Chikahiro, born 1976, had left Keio University in 1999 for Mitsui & Co., where he was posted to the information-industry division, spent time in the United States helping foreign startups build businesses in Japan, and on his return ran internal ventures and oversaw subsidiaries — a career spent watching new businesses come into being from the supporting side. The material he chose when he finally left was the business card. Some ten billion of them are said to change hands worldwide, and almost none of that information becomes anyone's asset: it accumulates in an individual's drawer while the company cannot say who met whom, or when. The bet was not on the size of a market but on the breadth of a neglected inconvenience.

The first service, Link Knowledge, launched three months after incorporation, in September 2007: scan a card and the name, company, title and email address land in the cloud, visible to the whole firm, so that the contact survives the transfer or resignation of whoever made it. The method behind it looked, for its moment, distinctly backward. Sansan refused to leave the reading to machines and put human operators into the loop, reaching 99.9% accuracy. Terada explained the choice by the nature of accuracy itself: 98% is fine for translation, but an email address with one character wrong is a dead record. So build accurate data by hand first, then hand it to an AI engine once enough of it exists — accuracy fixed first, technology brought to it afterwards. Eighteen years of digitizing analogue information, he said in 2025, is precisely why the company can now use AI well; he is careful to add that it was never designed as a moat, only as the appropriate answer to the problem.

The first six years were profitable and small. Cloud-delivered software was itself a rarity in Japan in 2007, and the model of raising money to fund losses on the way to scale was legible neither to investors nor to customers, so Sansan put in its own money and held down spending. It turned profits. It did not accelerate. Outside capital raised across those six years came to $626,645 (¥50m). Run carefully, the business would not die; it would also never move a commercial custom as entrenched as the exchange of cards. Around 2010 Terada judged his own company, in his word, puny. Before the turn came, he widened the door: Eight, launched in February 2012, opened the same machinery to individuals for free, reaching users through one salesperson rather than a corporate budget — and for years produced nothing but cost.

$626,645 (¥50m) in six years; then roughly $90.6M (¥10bn) in five. A private placement of about $5.1M (¥500m) in April 2013 changed the shape of the balance sheet, and the rounds followed in sequence — ¥1.4bn in May 2014, ¥2.0bn in January 2016, ¥4.2bn in July 2017, ¥3.0bn in December 2018. The founder's stake was diluted at every one of them, and accepting that dilution was part of the decision rather than a side effect of it.

The money went into making a market, not into chasing sales. Revenue for the year to May 2018 was ¥7.32bn against selling and administrative expenses of ¥8.95bn — an operating loss of ¥3.06bn; the following year revenue reached ¥10.2bn and the loss narrowed to ¥850m. In a software business with gross margins above 80%, a loss of that size is not a condition but a decision, since almost every yen of cost is one management chose to spend. What separated Sansan from the many companies that made the identical argument and vanished was the recovery mechanism sitting behind it: trailing twelve-month churn had fallen to 0.66% by the year to May 2019, low enough that the company said the point was to hold it under 1% rather than push it lower. Spend early and the revenue does compound — but only if that number holds.

In August 2013 the service was renamed from Link Knowledge to Sansan, and in the same month the first television commercial went on air. Mass advertising for business software was close to unheard of, because the buyer is a manager rather than a viewer and the return is hard to argue. What the advertising sold, though, was not a product but a classification. Sharing business cards in the cloud had no name, and without a name there was no budget line and no language for an internal approval memo; television told the country that a category called business card management existed, and let a manager explain it upstairs. The company then followed its own product: in March 2014 it moved to Jingumae and changed its name from Sanzan K.K. to Sansan, Inc. It listed on the Tokyo Stock Exchange's Mothers market on 19 June 2019 at an offer price of $41 (¥4,500), raising some ¥2.1bn and passing ¥100bn in market value, with Terada holding 37.1%. At that point it had over 5,800 corporate contracts and covered roughly 1% of Japan's workforce — room, as he put it, to multiply users a hundredfold — while the Singapore subsidiary opened in 2015 had produced nothing worth reporting.

In May 2020 Sansan launched Bill One, which receives invoices in the cloud and turns them into data — the human-plus-AI pipeline built for business cards, moved intact onto a different piece of paper. Paper invoices, Terada argued, are like business cards in the only way that matters: indispensable, and yet nobody treats them as a problem. The launch fell in the month after Japan's first state of emergency, which made paper you can only receive by going to the office into a universal grievance overnight.

The same conditions reached the core business as a headwind. Cards digitized fell 65.1% year on year in the week of 11 May 2020, recovering to a 13.6% decline by late June. Revenue did not break, because pricing was an annual contract on a pre-agreed volume rather than a charge per card exchanged; the older plans were annual per user. The structural weakness of business cards is that they depend on people meeting, and that weakness had already been removed on the pricing side, for other reasons, before anyone knew it would be tested.

Two tailwinds then stacked. Restricted attendance at offices made the inconvenience visible, and the revised electronic bookkeeping law, in force with a grace period from January 2022, put a deadline on doing something about it. Bill One reached ¥1bn in annual recurring revenue eighteen months from launch, which Terada thought was probably the fastest in Japanese SaaS; in that same January the third sheet of paper arrived as Contract One. Reading this as luck would be inaccurate — the invoice was chosen two years before the law changed, and both the law and the pandemic came afterwards. In 2022 the tagline for the Sansan service became “the database that makes sales strong,” the first time since founding that the words business card management left the product description. Dropping the label was possible only because the position was held: an 83% share of corporate business-card management and 8,186 contracts as of November 2021.

Revenue for the year to May 2025 was $288.7M (¥43bn) with an operating profit of ¥2.8bn — twenty-two times the ¥1.96bn of a decade earlier — and headcount had gone from 402 in the year to May 2018 to 2,235, at an average age of 31.7 and average tenure of 3.1 years. Eight, which had lost money since 2012, finally cleared: a first full-year adjusted segment profit of ¥314m in the year to May 2024, and ¥63m after corporate cost allocation the year after. Twelve years from launch.

The second engine repeated the first pattern with one difference. Consolidated operating profit fell from ¥757m in the year to May 2020 to ¥199m in the year to May 2023 as advertising and headcount were concentrated on Bill One — but the core had not weakened, the Sansan segment running a 51.0% operating margin in the year to May 2022 and 56.0% adjusted the year after. Group profit looked thin only by the amount being moved from the earning side to the next business. What differed from 2013 was that an existing business was already profitable and that no new shares were issued: the first ¥10bn came from repeated placements, the money behind Bill One from the IPO proceeds and the cash the core threw off. Acquisitions filled in the reading-and-parsing stack around it — the transcription firm Logmi in 2020, the natural-language house Language Understanding Research in 2023, sites in the Philippines in 2023 and Thailand in 2024 — with the 2023 purchase of Creative Survey the exception, a tool for collecting customer voices rather than digitizing paper.

In May 2025 the capital and business alliance with Unipos was unwound, with ¥2.30bn taken as an extraordinary loss on the share sale — a partner Terada had spoken of in 2022 as one whose affinity with Sansan would grow. In the year to May 2026 the shape of hiring itself changed: an organization built around AI use let the company slow recruitment and cut its adjusted SG&A ratio by 6.2 points, it sold Logmi, and it guided a first dividend of ¥5.00 a share for the year to May 2027 — thirteen years after deciding to dig the hole, returning cash. And the sharpest statement of the risk comes from the founder. Asked in October 2025 what the greatest one was, he named not competition but AI: that people may come to see AI as having swallowed Sansan and rendered its interface meaningless. The company's strength has always assumed that contact between businesses leaves an analogue trace — and he had already said in 2022 that roughly three in ten encounters had moved online for good. Whether Sansan can earn revenue that does not depend on the number of sheets left to read is the question of its next decade.