Apache Ossie joins incubator to standardise metrics

Open Semantic Interchange has entered the Apache Incubator under the new name Apache Ossie (Incubating), placing the project under the Apache Software Foundation.

The project is an open specification designed to standardise how companies define business metrics, dimensions and relationships across software tools and data systems. Its backers say common business terms such as Monthly Active Users are often defined differently across warehouses, customer systems and business intelligence platforms, producing inconsistent results for analysts and software agents.

The name change from Open Semantic Interchange, or OSI, was made to avoid confusion with other open source projects that use the same acronym. The underlying specification remains unchanged and continues to use a YAML-based format for semantic definitions.

Broader backing

Support for the initiative has expanded since its launch last year. The coalition has grown from 17 founding partners to more than 50 organisations, including Databricks, Salesforce, Oracle, dbt Labs, Informatica, Collibra, Qlik and BlackRock.

Activity around the project has also increased since the code repository opened in November 2025. Project details show that contributors from Snowflake, Salesforce, Databricks, dbt Labs, RelationalAI, GoodData and Honeydew have made more than 100 commits and 35 merged pull requests.

Three working groups cover metric language, catalog and ontology. Reference implementations include a converter for dbt MetricFlow, a converter for Apache Polaris catalog metadata and a Snowflake semantic model converter.

Governance shift

Under the Apache Software Foundation, the project will follow public development processes and consensus-based decision making. Changes to the specification will go through discussion and voting, and committer status will be based on contribution rather than employer.

The project's mentors include Apache figures who helped guide Apache Iceberg and Apache Polaris through the same incubation process. Supporters argue that this governance model is important for a specification intended to work across competing software vendors and data platforms.

Ossie is intended to serve as a shared semantic framework across the modern data stack. The goal is to let business intelligence tools, query engines and artificial intelligence agents use the same machine-readable business definitions without losing meaning as those definitions move between systems.

That matters because semantic inconsistency has become a practical problem as companies spread data and analytics work across multiple products. A metric defined one way in a warehouse and another way in a dashboard can produce different answers from the same underlying data. Artificial intelligence tools that rely on those definitions may also draw conclusions from conflicting business logic.

Snowflake role

Snowflake is a founding member and remains an active contributor. Several of its engineers are listed as initial committers and members of the project management committee for the incubating project.

Snowflake has tied the open specification to its own semantic products, including Semantic Views, Semantic View Autopilot and Horizon Context. It argues that those tools apply the same principles by using shared business definitions across analytics and artificial intelligence workflows inside its platform.

Still, under Apache governance, Ossie's direction is expected to be set by the wider community rather than any one vendor. Areas under discussion include broader support for complex metric logic and relationships, more converters for external platforms, a standard semantic query specification and integration with Apache Polaris so semantic models can be discovered through catalog systems.

Backers say that if adoption continues to widen, Ossie could become a common format for exchanging semantic definitions between software products that currently interpret business logic differently. For companies trying to make artificial intelligence systems, reporting tools and query engines work from the same definitions, that would address a long-running source of inconsistency in analytics.