Last month, a solar developer was planning a greenfield development in a new county. But the county’s commissioners were split on solar: one consistent champion, one real critic, and a swing vote that often decided approval.

Our client needed to know exactly what to say to each commissioner. What causes did they care about? Why did they vote the way they did? And where might pushback come from?

That used to take hours of research. Now, it can be done in minutes, thanks to Spark’s new Stakeholder Intelligence.

How it works

1. Search any county, township, or parcel. Spark will automatically collect all historical decisions, relevant board members, and meeting video recordings, both from our live database and using an active web scrape.

   

2. Click between commissions. Each location may have several authorities. Spark provides every relevant decision-making body with a full list of members.

3. Receive the high-level overview. See the historical voting record, the recent trendline, and a summary of what gets projects approved — and what kills them.

   

4. Click into any decision-maker for the full dossier. Spark supplements each commissioner’s voting record with detailed information, including their personal and political background, the causes they care about, and contact information.

5. Trace every item back to its source. As always, Spark is committed to keeping you in the driver’s seat: Every vote, every claim, and every detail is linked directly back to an online source so that you can dig deeper.

Behind the scenes

Beyond just scraping or querying ChatGPT, this feature is built with the depth to be the equivalent of reading thousands of pages and watching hours of meeting videos:

1. Finding the record. Every county publishes differently, so we search the open web for minutes and agendas, pull ordinance text, and separately hunt down each jurisdiction's own video channels.

2. Reading it. We pull from both meeting minutes (PDFs) and the videos. We convert minutes page-by-page with OCR when transcription is necessary, so any claim can be traced back to the exact page it came from.

3. Extracting the decisions. A meeting record can be eighty pages of procedure wrapped around three decisions that matter. We split the text so motions and roll calls stay intact, then pull out each decision: what was decided, the vote count, and who voted which way.

Keeping hallucinations down by honestly showing gaps: When we can't tell, we let you know, rather than guessing or hallucinating results. Where the minutes recorded a tally without names, nobody gets credit, so per-member vote counts are minimums, not totals. Due to this, you may occasionally see unknowns in the record. That’s the system working, not failing.

Get started today!

Run any Spark report to start receiving the inside scoop today.

Want to learn more or have feedback? Reach out to us — we love to hear from developers