Rebundling the Modern Data Stack
Charles Chretien, co-founder of Prequel, on the modern data stack’s ROI problem
This reveals that modern data tools are no longer selling a single step in the pipeline, they are racing to own a broader workflow before that step gets absorbed by a neighbor. Once warehouses became the center of gravity, customers could mix Fivetran for ingest, dbt for transforms, and Census or Hightouch for activation through SQL, which made each point tool easier to swap unless it added adjacent products and captured more of the budget.
-
Fivetran started as a connector business, moving data from SaaS apps and databases into Snowflake, BigQuery, and Redshift. That core job is valuable but exposed. Warehouses can add native replication, and SaaS vendors can ship their own warehouse connectors, which pressures Fivetran to move beyond raw ingest into transformation and other layers.
-
The unbundled warehouse stack made specialization possible in the first place. Segment once bundled collection, identity, and routing in one system. Then warehouse native tools split that stack into separate layers, with Fivetran on ingest, dbt on transform, and Census on activation. The current rebundling cycle is those layers trying to stitch themselves back together.
-
This is also about ROI. Add on tools that only process data are harder to defend in a tighter budget environment than products tied to a visible business outcome. That is why vertical data products like CDPs, which can point to better ad targeting or conversion lift, have a cleaner story than horizontal infrastructure alone.
The next phase is fewer standalone utilities and more data platforms that bundle ingest, transform, activation, and eventually compute or vertical workflows. The winners will be the tools that turn warehouse data into a direct business action, because that is the clearest way to stay embedded, grow spend, and avoid becoming a replaceable feature.