July 27 Dolt Bounty retro

https://www.dolthub.com/repositories/pdap/datasets/bounties/default

What went well?

• engagement & re-engagement with community + potential supporters

• iterated on our schema / planned and made some changes

• we filled out the agencies table to a high level of completion → a prerequisite to the datasets table

• excellent support from the Dolt team!

• learned that volunteer interest plays a big role in determining our focus

• got a start on refining the review process

• we split the bounty into two pieces so we had an opportunity to learn

What could have gone better?

• there's not currently an automated review process or script

• we discussed the difference between a dataset URL and an agency homepage

  • people ended up focusing on the agency homepage URL more than the datasets

• one person ran away with the scoreboard by using machine readable lat/long

Next steps

• Katie to share some metrics on engagement

• Minimize variables + types of submissions allowed, particularly differences in scraped vs. manually gathered data

• Plan next bounty for ~8/25 start

  • Formalize the fact that our team should do the reviews + scripts

  • Smaller bounties help us iterate on a faster loop

  • Look at attribution + scoreboard code and make contributions to how rewards are given out.

  • Announce $50/participant (pending Katie + Dolt approval)

Notes for bounty success

1. Pick a common schema for data (i.e. "incident reports")

2. Allow people to submit data linking back to the source (i.e. dataset)

3. Focus on discrete, scrapeable goals with relevant topics

   • e.g. hospital bounty was in response to a law which passed which required hospitals to publish data

   • national scope for a narrow data focus is a potential alternative to the narrow geo focus

     • helps with schema normalization + big stories + big moves

     • local is storytelling → action loop

4. Focus on additions vs edits, enforceable by requiring properties