If we leave the vendor, who owns the recordings, transcripts and scores?
You should, and the contract should say so in those words. Ownership and portability are different promises, though — a vendor can concede ownership while making export practically impossible. Ask for the export format, the mechanism, the window after termination, and whether derived outputs like scores, tags and summaries come with it. Request a sample export during evaluation, not during the breakup.
Ownership without portability is a word
Nearly every vendor will agree that your data is yours. Fewer will commit to how you get it back. The gap between those two positions is where switching costs live.
The distinction is not usually adversarial, either. Most vendors have simply never been asked, so no export path exists and building one competes with the roadmap. Asking during the sales cycle is the cheapest moment to get it built, because that is the only moment your leverage is high.
Portability is a set of concrete commitments: a documented format, a mechanism you can run without begging, a defined time window, and coverage of derived data as well as raw data. Ask for all four.
What a complete export contains
- Audio files with stable identifiers, not just a folder of timestamps.
- Transcripts with speaker labels and timestamps, joined to those identifiers.
- Structured outputs — scores, criteria, tags, outcomes — keyed to the same identifiers.
- Rubric and configuration definitions, so the scores are interpretable after you leave.
- Cross-system keys mapping records to your CRM or field service ids.
- Model and version stamps, so historical results can be read correctly.
The join keys are what usually goes missing
Exports tend to arrive as separate dumps with no shared identifier. You get audio files named by date, a transcript archive named differently, and a spreadsheet of scores with no way to connect them. Technically you received your data. Practically you received a puzzle.
Ask explicitly whether the export preserves the keys that link the three together, and whether it includes the mapping to your field service system's customer and job ids. That mapping is often the most valuable asset in the whole pipeline, because it is what took work to build.
Test the export in the first month
Request a sample export while the relationship is new and everyone is helpful. Load it somewhere and see whether you can reconstruct a report from it. Whatever is missing will be much easier to fix now than during a transition.
There is a design lesson behind this too. If results are written into a database you control rather than living only inside a vendor's product, the export question largely disappears — which is one reason integration-first architecture and reporting you own tend to age better than closed tools.
Topics: data ownership · portability · vendors · exit
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.