So you want to actually use your meeting data outside of Sembly's walls? Good luck. The export "functionality" is a classic example of vendor lock-in disguised as a feature.
Here's the reality of pushing Sembly data to Power BI:
* You get a JSON export. Not CSV, not a database dump. JSON. Hope your team is comfortable with Power Query's M language to parse nested structures.
* The API exists, but it's rate-limited and the documentation is... optimistic. Key metrics like speaker talk time are calculated fields they don't expose directly.
* The real cost isn't the subscription. It's the engineering hours to transform their schema into something analyzable.
Steps I had to take:
1. Use the API to pull meeting summaries and transcripts (separate calls).
2. Build a custom connector to flatten the JSON for participant-level data.
3. Re-calculate analytics they already show in the dashboard because the raw metrics aren't available.
You'll end up rebuilding half their reporting just to get a clean dataset. The promise of "insights" falls apart when you try to own the data.
Read the contract
Totally feel this. I actually tried it and the JSON export is a special kind of messy. My bigger issue? The schema changed twice in six months, broke my Power Query setup both times.
You mention re-calculating analytics - that was the final straw for me. If I'm paying for insights, I shouldn't have to manually rebuild the "engagement score" formula they won't share. Switched to a tool with a real SQL endpoint.
Demo or it didn't happen
Yeah, the JSON-to-M process in Power Query is a real grind. I built a function to recursively flatten their participant array, but it's brittle. My bigger headache was the timestamp formatting for speaker turns - they use a custom epoch-ish format that isn't documented.
You're spot on about rebuilding the analytics. I ended up reverse-engineering the "speaker talk time" from the transcript segments because the aggregated field wasn't in the API. Felt like I was paying them to do my data engineering.
Ever try pulling the data into a staging layer first? I started dumping the raw JSON into a Postgres table (with a JSONB column) and then using dbt to parse it out. At least when their schema shifts, I only have to fix the transformation logic, not the entire Power BI import.
Data is the new oil - but it's usually crude.