Digital Science is the company hiding in plain sight on a researcher’s browser. The scientist drafting in Overleaf may never think about the administrator tending a Symplectic Elements record, the librarian running a Figshare repository, or the funder searching Dimensions. Yet those products sit in the same London-based family. Together they cover the peculiar relay race of modern research: find money, find collaborators, write, publish, share data, count attention, report outcomes, defend the integrity of the whole affair.
That breadth is the point. A research paper is the visible artifact, but the work around it is a thicket of grant records, identifiers, spreadsheets, patents, datasets, compliance reviews and institutional reporting. Digital Science sells tools for that thicket. Its customers include individual researchers and students, but the larger contracts are institutional: universities, publishers, funders, government bodies, pharmaceutical companies and corporate R&D teams. Emory, USDA, Springer Nature, Novo Nordisk, Abbott and Sage appear among public customer examples. Indiana University has used Symplectic Elements for faculty activity reporting across a system serving nearly 8,000 faculty.
A publisher notices the rest of science
The company began in 2010 inside Macmillan, then home to Nature Publishing Group. Founding managing director Timo Hannay and his colleagues saw an obvious asymmetry: publishers served science at the publication stage, while researchers struggled through everything before and after it. Their answer was not another journal. It was a business that would back software across the research lifecycle.
The working title was Project Babbage, after computing pioneer Charles Babbage. Then somebody noticed the omen: Babbage was a genius whose grand machines remained unfinished and who died in financial disappointment. Digital Science sounded less like a warning from history.
More important than the name was a change in method. Rather than develop every tool centrally, the company would invest in founder-led specialists, help them with capital, distribution and industry access, and acquire some over time. The earliest launch-night paperwork tells the story neatly. While guests drank downstairs in December 2010, the co-founders of research-management company Symplectic were upstairs signing investment documents with Hannay and investor Sebastian Peck.
“Technology isn’t usually the challenge - the real challenge is culture change.”Daniel Hook, former CEO and Symplectic co-founder
That model still defines the group. Altmetric measures where research travels beyond citations. Figshare hosts and exposes research outputs. Overleaf gives technical authors a collaborative LaTeX editor. ReadCube manages literature. IFI CLAIMS supplies patent data. Writefull applies language models trained on scholarly writing. Metaphacts builds knowledge graphs. Symplectic records the institutional activity that must eventually be reported to somebody with a form.
The product that made the portfolio legible
A collection of good niche products can still be merely a collection. Dimensions, launched in its present form in 2018, made Digital Science’s portfolio thesis easier to see. Six group companies worked with more than 100 funders and universities to link grants, publications, citations, patents, clinical trials, datasets, policy documents and attention signals. Instead of asking only “Who cited this paper?” a user could follow money into research, research into inventions, and findings into policy.
The scale is now uncomfortably large for a footnote: more than 165 million publications, 8.2 million grants, 184 million patents, 74.8 million datasets and 1.2 million clinical trials. Those are Digital Science’s 2026 figures, and they move as records are added. Dimensions competes with Scopus and Web of Science, but its differentiation is the connective tissue across output types and a free personal search tier. Paid analytics, APIs and specialized applications handle the institutional work.
This is also how Digital Science makes money. There is no single all-you-can-eat price. It sells subscriptions, licensed datasets and APIs, hosted repositories, research-management software, premium individual plans, integration and specialist consulting. Institutional pricing is usually quote-based. Free access is a distribution layer, not the whole model. Its venture arm looks for aligned companies that are profitable or nearly so, while the Catalyst Grant makes much smaller, earlier bets - up to £25,000, equity-free, in the 2026 round.
A federation, not a feature buffet
Digital Science’s closest competitor is often a shopping list. A university might combine Clarivate or Elsevier analytics, a DSpace repository, EndNote or Zotero, spreadsheets and internal reporting systems. A lab might only need a reference manager. An author might happily use free Overleaf and never meet the parent company. Digital Science is different when the buyer needs several stages connected, institutional support and data that can survive an audit.
The portfolio structure has risks. Separate brands bring separate interfaces, histories and integration work. A suite can become a cabinet of excellent screwdrivers when the customer wanted a drill. Digital Science’s response has been to organize around shared hubs and connected data without erasing the products people chose in the first place. Its investment criteria are revealing: product fit, profitability and synergy. The company says its value should create a competitive advantage for the acquired business, rather than simply add another tile to a sales deck.
The model is most copyable under specific conditions. Start with a user community whose workflow has stubborn, specialist seams. Back founders with credibility in those seams. Share routes to market and data, but preserve product identity. Build the connective layer only after the individual tools earn trust. It works poorly when products have no common customers, no useful data to exchange, or margins too thin to support integrations, security and long institutional sales cycles.
Acquire domain expertise, not a logo. Keep the specialist team close to its users.
Offer a useful free edge while charging institutions for depth, scale and support.
A portfolio earns its keep when data and workflows cross product boundaries.
One small team with one job may be better served by one focused tool.
Now the database talks to agents
Digital Science had machine learning in products years before “AI” became mandatory in every quarterly presentation. Writefull, a 2016 Catalyst recipient, trained specialized language models on scholarly text; Digital Science invested in 2018 and bought the company outright in 2023. The company’s 2026 releases go further, but they are conspicuously preoccupied with control.
Papers AI, built by the team behind Overleaf, keeps drafts, references, datasets, Jupyter notebooks and task boards in one project. Its assistant sees that project context rather than one isolated prompt. Suggested changes remain reviewable. Users can connect an external or local model, and projects can stay on-device. The product has free and paid options. It is an attempt to fix a mundane research nuisance - ferrying context between tools - rather than automate the scientist out of the loop.
Dimensions and Altmetric also gained Model Context Protocol connectors in August 2026. These let licensed AI agents query live, structured research records and attention data inside tools such as Claude, ChatGPT and Gemini. A pharmaceutical analyst could map a drug class across papers, patents and trials; a university team could inspect funding relationships; a communications office could find research entering policy or clinical guidance. The sales argument is not that the model is magically wiser. It is that the model can show where its answer came from.
“Trust matters as much as capability here - researchers need to know they’re still the ones making the final call.”Juan Castro, principal AI scientist
That emphasis is practical. Research customers cannot shrug at a fabricated citation, a hidden edit or a compliance conclusion with no trail. Dimensions Research Strategy, announced in July, promises evidence-backed briefs for university leaders; the new Citation Check screens self-citation risk; Symplectic’s profile curation uses AI to reduce record-keeping. Each product aims at a bounded decision with provenance attached.
The baton passes, the thesis stays
On August 1, 2026, Stephen Leicht became chief executive. His biography fits the company’s method almost suspiciously well: Digital Science invested in ÜberResearch, which he co-founded, in 2013; that company’s grants data evolved into Dimensions. Daniel Hook, CEO for 11 years and a Symplectic co-founder, moved to parent company Holtzbrinck as chief scientific officer. The handover kept both men inside the wider system.
Digital Science now lists 501 to 1,000 employees and remains privately held, so its consolidated economics are not laid bare. What is visible is a company shifting its center of gravity. Universities and publishers remain foundational, while enterprise R&D, pharma, government, research security and commercialization are growing themes. A new Jisc agreement, for example, gives UK universities access to Dimensions Industry Partnerships to identify corporate collaborators and strengthen commercialization.
The durable idea is not that every research task belongs in one interface. It is that research creates a trail, and the trail becomes more valuable when its pieces can recognize one another. Digital Science spent 16 years collecting those pieces. In the agent era, the unfashionable work - identifiers, permissions, provenance, linked records and human review - may turn out to be the product.