- InfoWorld reports Microsoft and Google are joining Apache Ossie, an open standard for sharing business metric definitions like 'revenue' across analytics, AI and BI tools. It already has over 60 backers.
- It matters because lock-in now includes where business logic lives, and shared definitions could cut conflicting answers from AI agents. Analysts say the two backers do not guarantee lock-in will disappear.
- Ask your data team where your top metrics are defined and how hard they would be to move. Ask vendors for written commitments to export and import Ossie models.
The costly part of switching platforms is meaning
Moving data between analytics platforms is the easy part. Moving what the data means is harder. A metric like revenue carries rules. What counts? What is left out? How do the tables join? Those rules usually sit inside each vendor's tool. Teams rebuild them every time they adopt a new one.
Apache Ossie is an open effort to change that. It is a project of the Apache Software Foundation. It sets out a standard way to swap these definitions between analytics, AI and business intelligence tools. Its tagline is "Stop redefining 'Revenue' in every dashboard."
The lesson here is that lock-in is no longer only about where data is stored. It is also about where the business logic lives. A company that cannot move its definitions cannot easily move its tools.
Who is behind it
InfoWorld reports that Microsoft and Google are joining the project. It already has over 60 backers. Snowflake, Databricks, Oracle and Salesforce are among those InfoWorld names.
The effort began as Open Semantic Interchange. It was renamed Apache Ossie when the Apache Incubator accepted it in June. It is still in incubation.
How it works
Ossie describes a 'semantic model' in plain YAML or JSON files. Think of a semantic model as a shared dictionary. It tells each tool what a business term means. The specification has six building blocks: the model, datasets, fields, metrics, dimensions and relationships.
Datasets are business entities, such as fact and dimension tables. Fields are row-level attributes. Metrics are calculations such as sums, averages and ratios. They can span several datasets. Dimensions are categories like where, when and who. Relationships are the foreign keys that join datasets.
InfoWorld describes a hub-and-spoke design. Ossie is the hub, the common format. Converters link it to each vendor's own system. A company needs one converter per platform. Without a hub, it would need one for every pair of platforms. The project calls that tangle of custom links 'integration debt'.
What is committed, and what is not
Microsoft is building a converter that works in both directions. It would turn Power BI semantic models into Ossie files and back again. The company described this in a blog post, InfoWorld reports.
Microsoft is also pushing for Ossie to recognise DAX. That is the formula language Power BI uses for calculations. If it succeeds, the calculation logic could travel with the model.
Google's position is less settled. A company representative said Google is in the process of joining. It has not said what it will contribute. The specification already lists BigQuery and GoogleSQL as a supported dialect. The representative credited early community contributions.
Analysts urge caution. They told InfoWorld that the two backers make interoperable platforms a little more likely. They said it is no guarantee that lock-in will disappear. The benefits the project lists are its own claims. They are not measured results.
Why AI raises the stakes
A conflicting definition used to produce two numbers in two meetings. With AI agents, it can produce two answers from two agents. Stephanie Walter of HyperFrame Research said one shared definition could cut that risk. Agents would be less likely to read the same metric differently. That could give CIOs more confidence to scale agent use, she added.
Ashish Chaturvedi of HFS Research made a related point. A metric defined once becomes a versioned, reviewable piece of code. It no longer has to be rebuilt for each platform. For governance, the gain is review. Someone can inspect and approve the definition.
Questions to put to your data team
Ask where your top ten business metrics are defined today. Ask in how many tools. Ask how long it would take to move them to another platform.
Ask which of your vendors have committed to Ossie converters. Ask which only support the name. Treat this as a buying question. Ask for a written commitment to export and import Ossie models.
Ask who approves a change to the definition of revenue. Ask whether an AI agent would see that change.
Your data can be portable and still mean different things in every tool. Ossie is an attempt to make the meaning portable too. Its value will show up in the converters, not the announcements.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: Apache Software Foundation.
