Dentsu Soken: long-term performance & turning pointsSales (revenue) and profit-margin ratio
Sales (¥ bn)Net margin (%)
1975An agent between GE’s machines and Dentsu’s advertisers
Revenue (¥ bn, bars) · net margin (%, line)
Source: securities reports
1975Founded as a Dentsu–GE joint venture
1987US subsidiary; overseas network begins
1989Takes over Dentsu’s internal systems work
In December 1975 Dentsu and General Electric Information Services set up Information Services International-Dentsu (ISID) in Tokyo on a roughly even split. The business was to bring GE’s mainframe time-sharing service to Japan and sell it into the data-processing needs of Dentsu’s advertising clients — a company built entirely on borrowed technology and inherited customers, with the name spelling out the arrangement.
From the late 1980s it built the overseas network that suited that role: a London branch in 1986, a US subsidiary in 1987, Hong Kong in 1989, Asia in 1990, Europe in 1991, Singapore in 1992. The customers were Dentsu group companies and Dentsu’s advertisers, and the value offered was supporting the information systems of Japanese firms going abroad. In 1989 it began taking Dentsu’s own internal systems development and operations on a continuing basis — deepening the dependence on the parent, and securing an income floor that made everything else possible.
2000Listing, and the scramble not to become a subcontractor
Revenue (¥ bn, bars) · net margin (%, line)
Source: securities reports
FY2006 · unconsolidated
Revenue$591M
Net income$11M
Net margin1.9%
→
FY2007 · unconsolidated
Revenue$639M
Net income$14M
Net margin2.1%
2000Lists on the TSE First Section
2001Acquisition programme begins
2002¥72.4bn revenue; a ¥120bn target set for 2004
2004Head office moves to Minato
ISID listed on the First Section of the Tokyo Stock Exchange in November 2000, twenty-five years after founding and at the height of the IT boom. Both parents stayed on the register, but the binding constraint since 1975 — that anything significant required two shareholders to agree — was gone, and the company now had shares it could use as acquisition currency and as staff compensation, along with quarterly scrutiny it had never faced.
It used them immediately. Kisco Solution in 2001, a consulting joint venture with America’s International TechneGroup, SIID in 2002, a Shanghai company the same year — pursued through an explicit three-stage method: take a minority stake in a company with good technology and weak management, move to equity-method ownership above 20%, and go past 50% if it works. Against consolidated revenue of ¥72.4bn in the year to March 2002, management set a target of ¥120bn and ¥12bn of recurring profit by March 2004, growth of roughly 30% a year that assumed acquisitions would supply it. Trade press of the day grouped ISID with Fuji Soft ABC and TIS as the three mid-tier independents driving the M&A wave.
The urgency was defensive. Prime contracts on large systems projects were held by the manufacturer-affiliated giants — NTT Data and its peers — and ISID could field perhaps a tenth of their engineers, which pointed toward a future as a permanent subcontractor. Its answer was specialisation: expertise in advertising, marketing and content industries, plus the global footprint. In 2004 it moved its head office to Minato, where the IT industry clusters.
2008The founding business ends; a real integrator begins
Revenue (¥ bn, bars) · net margin (%, line)
Source: securities reports
FY2008 · unconsolidated
Revenue$748M
Net income$22M
Net margin3%
→
FY2019 · consolidated
Revenue$924M
Net income$57M
Net margin6.2%
2008GE exits; Dentsu becomes sole parent
2015Fiscal year end moved from March to December
2016Operating margin reaches 8.1%
2019¥100.7bn revenue, ¥10.1bn operating profit
After the 2008 crisis GE retreated to financial services, its information-services business was sold off, and its stake in ISID fell away until Dentsu was the sole parent. For a company created as a sales window for GE’s network, this was not the tidying-up of an unprofitable line: without becoming a company that made what it sold, no reason to exist remained. The technical base was rebuilt on open Linux and Windows systems, and the business was re-founded on three pillars — SAP and Oracle ERP implementation, PLM for manufacturers, and risk-management systems for financial institutions.
The rebuilt company then had to be tidied. Subsidiaries were merged or absorbed through 2009–2014, a consulting subsidiary was created in 2013, and engineering and Indonesian units were added. Revenue in the year to March 2013 was ¥72.8bn with ¥4.3bn of recurring profit — flat against a decade earlier — and ¥78.2bn by March 2015. In April 2015 the fiscal year end was moved from March to December, aligning with the American and European integrators and consultancies it wanted to be compared against, even though Dentsu itself stayed on March.
From there the margin arrived. Operating margin went from 3.7% in the transitional nine-month period to 8.1% in 2016, and revenue climbed from ¥79.8bn in 2016 to ¥83.4bn, ¥91.0bn and ¥100.7bn in 2019, with operating profit rising 55% over three years to ¥10.1bn. The company ran five segments — financial, manufacturing and business solutions, communication IT and cross-innovation — and kept adding joint ventures at the edges: with Fraunhofer in Germany in 2018, with Mitsubishi Estate on the FINOLAB fintech hub and with Seven Bank in 2019.
The pandemic turned corporate DX budgets into the company’s best market. Revenue went from ¥108.7bn in 2020 to ¥112.1bn and ¥129.1bn in 2022, with operating profit rising 84% over three years to ¥18.6bn and net profit doubling — the demand concentrated in ERP, cloud migration, data platforms and new digital services. It moved to the Prime market in April 2022 and to a board with an audit and supervisory committee in 2023, and held its growth as the cycle cooled: ¥142.6bn of revenue in 2023 and ¥152.6bn in 2024, each with ¥21.0bn of operating profit.
In January 2024 the company renamed itself Dentsu Soken, taking over the think tank of that name from Dentsu itself and absorbing its own consulting subsidiary; in April it bought Mitsue-Links, a large web production house, to bring UI and UX design in-house. The stated logic was that systems integration alone can no longer answer how complicated its clients’ problems have become, and that research, consulting, implementation and design should sit in one legal entity. The first year under the new name produced records: ¥164.8bn of revenue, ¥22.8bn of operating profit and ¥16.3bn of net profit in 2025.
What remains unsettled is ownership. The company that listed in 2000 to escape the veto of two parents is now, by press reports in 2026, the subject of a plan under which Dentsu keeps 61.8% and places the remainder with Fujitsu or a trading house. Fifty years after being founded to resell an American network to Japanese advertisers, its name says research institute and its share register says whose it is.
Choosing twenty-five years to be priced by the market
The effect of the listing cannot be measured by the ¥120bn target that was held up. Revenue was flat for several years, with loss-making periods in between. Even so, by going onto the public market this company came to hold shares usable both as consideration in acquisitions and as employee compensation, and became a company asked about its results every quarter. That a flotation once shelved in the early 1990s went ahead was because a single thing moved — the shareholders’ consent — rather than because the momentum of the business pushed it. Removing the constraint that had existed since founding, the wait for two parents to agree, can be seen as the real substance of this decision.
Twenty-six years on, the way that constraint was removed is being questioned in the opposite direction. According to reports in July 2026, the Dentsu Group is considering a plan to keep 61.8% while placing the remaining shares with Fujitsu or a trading company. It amounts to a design in which the parent’s judgement, supposedly diluted by the listing, once again covers the whole of the capital. A company that began in 1975 with Dentsu at 66% and GE at 34% settled, by way of the market, at Dentsu 61.8% — and a plan to leave the market is now under consideration. The worth of a decision to go public shows also in how the price is set when you come off.
When a company built on borrowed technology loses the lender
This company was born as an agent selling another firm’s equipment. Its role was to connect GE’s network with Dentsu’s customers, and the “International Information Services” in the corporate name amounts to a description of that role. The end of the founding business was therefore not merely the disposal of an unprofitable line. Unless it changed into a company that made what it sold, no reason for its existence would remain. The loss in the year to March 2003 can be seen as a figure produced in the middle of that reassembly.
The same thing happened on the capital side. Why the joint-venture partner left is not recorded in the disclosures, but for GE this company was the Japanese sales window for its own network, and once that network was folded the grounds for continuing to hold it disappeared as well. The Dentsu that remained raised its ratio past 60% and did not move it for the next twenty years. A company begun on borrowed technology loses its lender and settles inside a single parent — the 66-to-34 joint-venture ratio of 1975 has now been replaced by a different tug-of-war at 61.8 to 38.2.
A ¥150bn company adopting the sign of twenty-five people
The think tank taken over was an organisation of roughly twenty-five people, part-timers included. A company with revenue around ¥140bn adopted those twenty-five people’s name as its own. It is not an integration that can be explained by its contribution to profit; it amounts to changing the sign over the entrance in order to change where orders come in. Systems development is work you are called into after the specification has been decided, and how much you can charge depends on whether you can join the discussion before that. Placing recommendation, design and implementation in a single company can be seen as an attempt to reposition the firm at the top of that sequence.
That the name was taken from the parent group itself, however, gives this decision another meaning. A company that spoke at its 2000 listing of how the image of being a Dentsu subsidiary had run ahead of it took on, twenty-four years later, the name of Dentsu’s research institute. And two years after the change of name, the parent is considering a plan to place the remaining shares of just under 40% with Fujitsu or a trading house. The name Dentsu Soken, obtained by discarding the abbreviation ISID, is for now both a sign advertising social research and a nameplate indicating where the capital sits.
Each heading links to the full Japanese analysis — background, decision and outcome, with sources.
This is a condensed English edition. The full, source-by-source history — with the detailed narrative, financial tables, shareholders and executives — is maintained in Japanese: 日本語版（詳細）— Dentsu Soken full history in Japanese →
Dentsu Soken Inc. — 有価証券報告書 (annual securities reports) and quarterly results.
Shukan Toyo Keizai — 週刊東洋経済, 11 Jan 1997 (Japan–Korea–China joint survey by the original Dentsu Soken think tank).
Shukan Toyo Keizai — 週刊東洋経済, 7 Sep 2002 (industry report on the M&A strategies of the mid-tier IT integrators).
Press reports, July 2026, on the Dentsu Group’s plan for the remaining shareholding.
This page is provided for general information only and is not investment advice, nor a recommendation to buy or sell any security.
Figures are compiled independently and include our own estimates, approximations and machine-processed data; we make no warranty as to their accuracy or completeness.
Sources are primarily each company’s securities reports and other public filings, but errors and omissions may remain.
Any use of this information is at the reader’s own risk. Past performance does not indicate future results.
Company names, logos and other marks belong to their respective owners.
Data API
Dentsu Soken’s history, financials, executives and
shareholders are published as static JSON — no key, plain GET.