Transform was the headless metrics startup founded in 2020 by Nick Handel, Paul Yang, and James Mayfield, three engineers from Airbnb's data platform team. It is one of the most consequential semantic layer companies that no longer exists as an independent product, because the engine they built — MetricFlow — is now the technical foundation of the dbt Semantic Layer after dbt Labs acquired the company in February 2023.

The Transform story is also the story of how the entire metrics-store category briefly looked like the future of data, then collapsed into the larger dbt-shaped vortex.

The Airbnb Backstory: Minerva

To understand Transform, you have to understand Minerva. Around 2019, Airbnb's data team published an internal-then-eventually-blogged-about project called Minerva: a centralized metrics platform that defined every metric in Airbnb (bookings, GBV, ADR, host quality scores, etc.) once, in code, with full lineage and consistency guarantees. Minerva was built because Airbnb had been bitten repeatedly by the same problem every fast-growing data org hits: different teams used different definitions of "active host," dashboards disagreed, and the company spent meeting hours arguing about whose number was right.

Minerva fixed this internally. The engineers who built Minerva realized the same problem existed at every company larger than 100 people and saw a venture-backable opportunity. Nick Handel had been the product manager for Minerva. Paul Yang had been an engineer on the same project. They left Airbnb in 2020 to found Transform with the explicit goal of commercializing Minerva for everyone else.

Transform raised seed and Series A funding from Index Ventures and Redpoint, hired aggressively, and shipped a product that exposed metric definitions as a service. The pitch was clean: "Minerva-as-a-service." For about eighteen months in 2021-2022, the metrics-store category was one of the hottest sub-segments in the modern data stack, with Transform, Supergrain, Cube, Trace, and a few others all chasing the same vision.

Note on Supergrain

Supergrain was a separate startup, founded around 2020 by ex-Stripe engineers, that was also pursuing the metrics layer. It was later acquired by dbt Labs as well, though the more consequential acquisition is widely understood to be Transform / MetricFlow. The two companies are sometimes conflated in old blog posts, hence the combined page title here, but historically they were distinct teams pursuing parallel approaches.

What MetricFlow Is

MetricFlow is the open-source query engine Transform built. It is, more or less, the runtime that Minerva's design implies: a system that takes a metric definition (in YAML) and a metric request (in a query API) and compiles the result into optimized SQL against a warehouse.

MetricFlow's design choices reflect lessons from Minerva:

• Semantic models are first-class. A semantic model declares an entity (orders, users, sessions), its dimensions, and its measures. Metrics are then defined on top of semantic models, not directly against tables.
• Joins are inferred. If two semantic models share an entity (both have a user_id), MetricFlow knows it can join them and resolves the join graph automatically when a query asks for a metric that spans them.
• Multiple metric types. Simple, ratio, derived, cumulative, conversion — MetricFlow supports a richer set of metric patterns than a hand-written SQL view could express cleanly.
• Compiled SQL. MetricFlow does not store data. It writes SQL against the warehouse, which means the metrics inherit warehouse performance, security, and access control.

This was technically much more sophisticated than dbt's original v1 metrics block, which is exactly why dbt Labs wanted it.

The Acquisition

dbt Labs acquired Transform in February 2023. The price was not disclosed. The strategic logic was obvious to everyone watching the category:

• dbt's v1 metrics layer was technically inadequate and customers were complaining.
• dbt Labs needed a real semantic layer to compete with LookML and Cube.
• Transform had the engine, the team, and the design pattern that fit dbt's worldview.
• Buying a small Series A startup was cheaper and faster than building from scratch.

After the acquisition, Transform's standalone product was deprecated. Existing customers were migrated to dbt Cloud's Semantic Layer. The MetricFlow open-source project was moved into the dbt Labs GitHub org and is now maintained as the engine behind the dbt Semantic Layer. Nick Handel joined dbt Labs as Director of Product for the Semantic Layer.

The Opinionated Take

The Transform acquisition is one of the cleanest "build vs buy" wins in modern data infrastructure. dbt Labs got a production-grade metric engine and a respected founding team for a fraction of what it would have cost to build internally. Transform's investors got an exit. The customers who had bet on Transform got migrated to a much larger and more durable platform.

The losers, arguably, were the metrics-store category as an independent segment. Once dbt Labs had MetricFlow, the case for buying a standalone metrics store from a Series A startup got dramatically weaker. Within a year of the Transform acquisition, the other independent metrics startups had either pivoted, been acquired, or shut down. The category collapsed into "dbt Semantic Layer vs Cube vs LookML," and Transform stopped existing as a brand.

The deeper lesson: the semantic layer wants to live next to the transformation layer. Standalone metrics stores fight gravity. Once the same team is writing the dbt models that produce the data, the same team wants to write the metric definitions that consume the data, in the same repo, in the same language. dbt Labs understood this. Transform's acquisition was the moment the rest of the industry was forced to agree.

Where MetricFlow Fits Today

MetricFlow is no longer a standalone product. It is:

• The query engine inside the dbt Semantic Layer
• An open-source project (BSL license) you can run locally for development
• The reference implementation of the dbt-style metrics design pattern

If you want to use MetricFlow in production with a real query API, you use dbt Cloud's Semantic Layer.