Scaling sentiment: Using digital surveys to geotarget vaccine distribution
summary
California deploys surveys on 12 state websites and collects 90,000 responses (including ~80,000 open text answers) then uses SurveyMonkey, Google Cloud/BigQuery, NLP, topic modeling and linear regression to model hesitancy at the zip code level; models stabilize in about four to six weeks and data products feed ad buys, outreach and logistics. Result: within 30 days 70% of pop-up/mobile doses target HPI quartiles 1–2 and by August those quartiles show 57% higher vaccination rates.
AI-written summary · about
chapters (10)
- 0:00Introduction and presenters
- 3:31Project overview and background
- 7:46Gathering sentiment data at scale
- 12:20Applying rigorous methodology
- 16:53Structured and unstructured data insights
- 21:31Equity lens and data overlay
- 26:12Zip code-level modeling and analysis
- 29:05Making data actionable for stakeholders
- 31:53Impact on vaccination equity
- 33:58Q&A session
related talks
29mHow data science shortened the COVID-19 pandemic by 2 months
01h 01mReal-time surveillance of potential epidemics
01h 07mPolitical behavioural targeting
27mData storytelling in the era of generative AI
42mVisualizing COVID-19 Data
01h 15mSunday Keynotes
20mFrom Data to Development: Using Stats To Keep Up With Mobile Game Trends
22mCOVID-19: Data Science & Expertise