Market Opportunities and Assessment Application

Inspiration

An automotive manufacturer operates in 25 states but has no data-driven framework to evaluate the remaining 26 for expansion. We set out to combine internal sales data with Census demographics to determine which regions fit our buyer profile, what the risks are, and whether the bigger opportunity is new markets or deepening existing ones.

What it does

The application scores all 51 US states on demographic fit, expansion risk, and revenue potential, while segmenting the country into 9 marketing personas with AI-powered sentiment analysis from 30,000 warranty claims. A 6-page interactive Streamlit app deployed on Snowflake serves as both a CEO executive summary and a deep-dive analytical tool.

How we built it

We joined an internal data warehouse (100K sales, 30K warranty claims, 1,500 dealers) with Census data from the Snowflake Marketplace, built 10 analytical views for fit scoring, risk assessment, and revenue projections, and used Cortex AI_SENTIMENT for customer sentiment mapping. The Streamlit app was developed locally and deployed to Snowflake — all in a single hackathon session using Cortex Code.

Challenges we ran into

Sales data was uniformly distributed (~4,000 per state regardless of population), which meant the model could rank states at a macro level but not reliably recommend specific vehicle segments. We also had to rebuild the risk model after the first version was dominated by existing market presence rather than actual expansion risk.

Accomplishments that we're proud of

We ran an honest self-audit that identified weak signals — product recommendations and large-state revenue projections — and removed them from the app rather than presenting noise as insight. The most useful finding was unexpected: 7 states are saturated while large states like CA, TX, and FL are at 15-28% of the penetration benchmark, suggesting deepening existing markets may outweigh new expansion.

What we learned

Data distribution matters more than model sophistication — uniform sales data limits even a well-designed model to "this state looks like our average market." Separating existing-market analysis from expansion analysis, and the Snowflake ecosystem (Marketplace, Cortex AI, Streamlit-in-Snowflake), were the two decisions that made the project work.

What's next for Market Opportunities and Assessment Application

1. Plug in real sales data — genuine geographic variation would unlock product-level recommendations per market.
2. Add competitive landscape data — overlay competitor dealer footprints and market share to identify truly open markets.
3. Move to sub-state granularity — ZIP or MSA-level analysis for the top expansion targets to enable precise go-to-market planning.
4. Model dealer network buildout costs — turn the scoring model into an actionable business case with capital and timeline estimates.
5. Upgrade the Census data — the paid ACS 2020-2024 dataset with 63 estimates would enable income-segmented demographic profiling.

Built With

• coco
• python
• snowflake-ai
• streamlit