Stat ClubFall 2026 Data Challenge
Food Need vs. Demand
Where Minnesotans need food help, where they actually ask for it, and why the two don't line up.
Teams of up to 4
The Client
The Food Group
Teams work with real data from The Food Group, a Minnesota hunger-relief organization that runs a food bank, a HelpLine connecting callers to SNAP and food resources, and MarketBucks, which matches SNAP dollars spent at farmers markets. Treat them as a real client: every result should connect to a decision their staff could make.
The Core Idea
Need is the help a household must have. Demand is the help a household actually asks for. A food shelf visit measures demand: some people in need never show up (stigma, no car, never heard of it), and some people who show up aren't food insecure.
Because resources follow visits, they follow demand. Easy-to-reach areas draw more food and funding, while high-need areas that are hard to reach draw little, and the data gives no warning. Your findings feed four real decisions: how to allocate resources, where new food shelves are needed, how to argue for a policy, and how The Food Group measures its own results.
Two Tracks, One Dataset
Analysis track
Answer four questions about need vs. demand with statistics you can defend.
- Judged by a panel against a rubric
- Deliverable: 5–7 minute talk, slides, and code
Prediction track
Forecast July 2026 food shelf visits, pounds distributed, and individuals served for the 86 Minnesota counties with food shelves.
- Scored automatically (RMSLE) against held-out data
- Deliverable: one predictions.csv
Every team does both tracks. Overall ranking is 80% analysis + 20% prediction, with a separate award for the best forecast.
The Questions
Q1
Do need estimates match food shelf activity?
Compare Feeding America's Map the Meal Gap county estimates with food shelf activity. Which counties break the pattern, and what explains the mismatch?
Q4
Do grocery prices drive food shelf traffic?
When grocery prices rise, do more people come to food shelves — right away, or after a delay?
Q5
SNAP and food shelves: partners or substitutes?
Do SNAP enrollment and food shelf use move together, or does one replace the other? What happens when SNAP benefits change?
Q6
Build a measure of hidden need
Create a new metric, using public data beyond demand, that flags places where need is probably bigger than the activity data shows.
How You're Judged
| Analysis criterion | Weight |
|---|---|
| Q1, Q4, Q5: complete, insightful, actionable | 30% |
| Appropriate, correctly applied methods | 25% |
| Q6: a measure of need that goes beyond demand | 15% |
| Outside data: well chosen and properly cited | 10% |
| Creativity and value to The Food Group | 10% |
| Communication, with every member engaged | 10% |
Each analysis criterion is scored 1–4; the rubric is adapted from MinneMUDAC's. Undergrad methods are enough to win — a method you can explain beats a complex one.
What You Hand In
- predictions.csv for the prediction track (due Sat, Oct 31, 11:59 PM)
- Code in your team's private repo that reproduces your predictions and every number in your slides (due Sat, Oct 31, 11:59 PM)
- Slide deck (PDF) answering Q1, Q4, Q5, and Q6, emailed to uwstatclub@gmail.com (due Thu, Nov 5, 3:00 PM)
- 5–7 minute presentation plus Q&A with judges (Thu, Nov 5, 6:00 PM). Every member should be ready to present and answer questions.
How to Submit
- You'll need a GitHub account: the registration form asks for each member's username.
- After registration closes, each team gets a private GitHub repo with the data and a predictions.csv template. Only your team and the board can see it.
- Your submission is whatever is on main at 11:59 PM Sat, Oct 31. There's nothing to upload: just push. At the deadline your repo locks to read-only.
- Every push runs an automatic format check on predictions.csv (✅ or ❌). It checks that the file can be scored, not that it's accurate.
- We rerun your code. If it can't reproduce your predictions.csv, you may lose your prediction score.
- Email your slides (PDF) to uwstatclub@gmail.com by 3:00 PM Thu, Nov 5, with your team name in the subject line.
Timeline
- Thu, Oct 1, 2026
Data Challenge Kickoff
The Fall 2026 challenge, Food Need vs. Demand, officially begins.
- Thu, Oct 8, 2026
Workshop: Graduating from Notebooks
Presented by Spencer Venancio. Move your analysis out of notebooks and into source files, a clean project structure, Git, and GitHub. Follow along in Python or R.
- Fri, Oct 9, 2026
Team registration closes
- Sat, Oct 31, 2026
predictions.csv and code due
- Thu, Nov 5, 2026
Slides due
- Thu, Nov 5, 2026
Data Challenge Final Presentations
Ground Rules
The data is not public
The Food Group's files are internal. Keep them in the club's private repo, never on public GitHub, Kaggle, or an open link. Findings can go in portfolios; raw data can't.
Forecast with repo files only
Your forecast may use only the files in your team's repo, which stop at June 30, 2026: no outside data. Breaking this rule scores 0 on the prediction track.
Free, public outside data only
Anything you had to buy, or got privately, disqualifies your team. Map the Meal Gap is required — request it from Feeding America early, since approval takes time.
Report anything personal
Names and phone numbers were stripped out. If you spot anything that identifies a person, stop and tell the board right away.
AI is allowed, with a catch
"The AI suggested it" is not a reason. Know where every number came from.
Questions?
Ask in our Discord or email us at uwstatclub@gmail.com.