The Associated Press had a peculiar distribution problem. It could deliver the news, yet the data behind a story could still end up with the wrong person in a newsroom. A spreadsheet might be difficult to find, awkward to connect to an analysis tool, or simply stranded outside the reporter’s working day. Information had arrived. Understanding was running late.
- The job: make enterprise data findable, understandable and governed.
- The method: connect datasets to definitions, people, policies and their history.
- The evidence: practical deployments at AP and WPP, with customer-reported gains.
- The turn: ServiceNow completed its acquisition in July 2025.
That is a useful place to begin with data.world. Enterprise software names tend to promise either heroic power or dreary housekeeping. This one sounds like an invitation. Its business rests on an unglamorous observation: a company can accumulate information faster than its employees can work out what any of it means.
01 / The spreadsheet had the wrong audience
AP’s early case study describes an organization spending significant resources on infrastructure for its data services while members struggled to make use of the results. The first failure was delivery and usability. Having good data did not ensure it reached someone who could turn it into a local story.
With data.world, AP created a shared place for vetted datasets and collaboration. More than 300 organizations could access it. The published case study reports doubled data production and more than doubled customer reach. Those are AP case-study results, rather than a promise that any newsroom installing the software will get the same arithmetic.
The practical change was modest enough to copy: keep the material, the explanation and the conversation together. A local reporter should not need to retrace the national desk’s entire research process before asking a local question. The value of distribution increases when the recipient can understand and reuse what arrives.
02 / A library with the conversations left in
A data catalog is an inventory of information assets, with enough description to make them useful. data.world sells a cloud-native version for enterprises. Analysts can search for resources; stewards can manage definitions and governance; teams can record who owns an asset and how it relates to other work. The catalog gives an organization a place to look before sending another hopeful message to the person who might know.
Its distinguishing architectural choice is a knowledge graph. Think of a dataset connected to a business term, an owner, a policy and a downstream report. The connections are themselves part of the information. Looking up a table can lead to the meaning of a metric, the person responsible for it, or the work that depends on it.
That approach places data.world in the market for enterprise metadata management, discovery and governance. Buyers also encounter Alation, Collibra, Atlan and Informatica. The sensible comparison concerns the systems a catalog can connect to, the governance it supports and whether ordinary colleagues will use it. A beautiful inventory with no visitors has the commercial appeal of a restaurant with excellent filing cabinets.
data.world’s expertise sits in semantic technology, metadata and collaboration. Its developer ecosystem includes a Python package for working with datasets and APIs, alongside an R library and a JDBC driver. The interface matters, but so does the ability to fit into existing analytical work rather than demanding that every task begin again elsewhere.
03 / The agency that cataloged its own memory
WPP provides a more expansive example. Its data.world deployment links data with code repositories, people and case studies. Someone finding a dataset can discover previous work around it and colleagues with relevant experience. A catalog becomes a route into organizational memory.
The company’s published case study reports more than 1,500 consistently active users and use across 70% of WPP agencies. Adoption began with cataloging existing material and showing people what they could do with it. The convincing demonstration was a useful resource found more easily.
“Getting our data out of silos and shared across teams is how we grow.”Vip Parmar / WPP global head of data
AP and WPP suggest the same buying test: choose an actual task and see whether the catalog removes a frustrating step. Does a reporter find a usable dataset? Does an agency employee find relevant prior work? Success needs a human verb. Counting imported assets alone leaves the important question unanswered.
04 / Four founders, one very technical owl
data.world was incorporated in September 2015, according to Brett Hurt’s fundraising retrospective, and launched publicly on July 11, 2016. Hurt had previously built Coremetrics and Bazaarvoice. His co-founders were Matt Laessig, Jon Loyens and Bryon Jacob, bringing experience from HomeAway and, for Laessig and Loyens, Bazaarvoice.
The original ambition extended beyond enterprise procurement. The public platform used linked-data technology and social features to help people discover, prepare and share datasets. By 2022, the company reported more than 1.6 million members in its open-data community. Membership measures the community’s reach; it does not tell us how many organizations bought subscriptions.
The mascot supplies a small engineering joke. The owl is called Sparkle: OWL refers to Web Ontology Language, while SPARQL is a Semantic Web query language. Hurt has described employees’ personalized owl avatars, or Sparkletars. Even a metadata company is entitled to fancy dress.
Its public-benefit identity also shaped the pitch. Company materials described data.world as a public benefit corporation and Certified B Corporation, pairing enterprise subscriptions with an open-data mission. In 2022 it reported six consecutive Austin Best Places to Work awards. These are historical descriptions of the independent company, and they help explain why the community remained part of its story as enterprise sales grew.
05 / Context acquired a capital budget
The idea took substantial financing. data.world disclosed $14 million at launch, another $18.7 million in 2017 and a $26 million round in September 2020. A $50 million Series C led by Goldman Sachs Asset Management followed in April 2022, bringing the company’s reported total fundraising to $132.3 million.
Cumulative totals reported in funding announcements. These are investment dollars, not revenue or customer savings.
For customers, the business model is subscription software. Enterprise pricing goes through a sales conversation that includes how many catalog users need access. The relevant budget therefore includes the subscription and the work of connecting systems, curating definitions and getting colleagues to participate. The published material does not justify inventing a universal project price.
Partnerships and integrations make that work less isolated. data.world joined Snowflake Partner Connect in 2020; that year’s funding announcement also named AWS, MANTA and Semantic Web Company. In 2023, it acquired Mighty Canary technology for a DataOps application that sends data-health updates and catalog context into the tools consumers already use. Its announcement offered the application without an additional charge to users on Standard and above.
06 / The next reader was a machine
In March 2024, data.world announced its AI Context Engine; full rollout followed in May. The product supplies organizational context to applications using large language models. The underlying problem is familiar from the newsroom: access to material does not settle its meaning. A machine interpreting internal data also needs the organization’s definitions and relationships.
Consider an illustrative question about retention. Does it mean customers retained, revenue retained, or a particular business unit’s measure? A plausible answer to the wrong definition remains wrong. The AI Context Engine uses the catalog’s knowledge graph to bring those connections into the interaction and support explanation and governance.
This explains the continuity in the company’s development better than a sudden conversion to AI does. Linked data mattered at the public launch. Business context mattered to the enterprise catalog. Generative AI gave those connections another audience. The company’s product direction changed as the potential user changed.
ServiceNow announced an agreement to acquire data.world in May 2025 and confirmed the July closing in its second-quarter results. It said the catalog and governance platform would strengthen AI agent understanding and enterprise data intelligence. Financial terms were undisclosed. The current data.world website describes the product as “now by ServiceNow.”
For a prospective buyer, the lesson is pleasingly mundane. Start with a recurring question, identify the people who can answer it and preserve the explanation where the next person will look. That is an editorial inference from the customer stories, rather than a guaranteed implementation recipe. It depends on maintained definitions, useful connections and participation. If nobody owns the meaning, a search box cannot negotiate agreement on everyone’s behalf.
The interesting thing about data.world is how stubbornly human its technical proposition remains. A number needs a definition. A dataset needs a history. A colleague needs a way to find both. Give an AI application the same assignment, and the old housekeeping suddenly has a very fashionable audience.
Follow the connections
Explore the platform, its developer tools and conversations with the people building it.