The Vendor Lock-In Audit: How Much of Your Data Stack Do You Actually Own?

Nobody chooses lock-in. It accumulates: a no-code ETL tool that was faster to start with, a BI platform where ten years of report logic now lives, a CDP whose export API conveniently doesn't include the fields you'd need to leave. Then one renewal cycle the price jumps 40%, and you discover the real product was the exit barrier.

We've sat on both sides of this — running vendor evaluations and RFPs inside enterprises, and engineering the escapes. Here's the audit we run.

The four flavors of lock-in

Data gravity: your raw history lives in their store, and bulk export is slow, fee-based, or lossy. Logic capture: years of transformations, metric definitions, and report logic encoded in a proprietary format that exports as, at best, a PDF. Workflow capture: your team's daily muscle memory and downstream systems are wired to their interfaces. Pricing capture: per-seat or usage-based pricing that scales with your success — the bill grows precisely when switching is most disruptive.

Most stacks have all four at once, and the vendor's account team knows your score better than you do.

The audit: three questions per tool

For each vendor in the data path, answer honestly: One — if we cancelled today, what data, logic, and workflows could we take with us, in what format, at what cost? Two — what would it cost to rebuild the irreplaceable parts on open components? Three — what's the five-year spend trajectory at our growth rate, including the repricing event the vendor will eventually attempt?

Score each tool: green (exportable, replaceable, fairly priced), yellow (one capture vector), red (multiple capture vectors and rising spend). The red list is your engineering roadmap.

The seam strategy

Full independence is the wrong goal — building your own BI tool is how data teams die. The right goal is owning the seams: your raw data lands in storage you control before any vendor touches it; your metric definitions and transformations live in version-controlled code (dbt, SQL, Python), not a vendor GUI; and every vendor in the stack is replaceable in a quarter, not a year.

Own the data model, the metrics layer, and the connectors to systems that differentiate you. Rent the commodity: dashboard rendering, standard SaaS ingestion, orchestration. Vendors are fine — captive customers are not.

Buying back ownership, incrementally

The escape is never a big bang. Land raw data in your own storage first — it's cheap and immediately defuses data gravity. Then port transformation logic to code, one subject area at a time, validating outputs against the vendor's. By the time you cancel, the cancellation is an anticlimax.

If you want a second pair of eyes on your stack, this audit is a two-week fixed-scope diagnostic for us — you get the scorecard and the exit costs whether or not you ever execute the plan.