Goodberry's Best Flavor 2026 — a real STAR election¶
▶ Live on BetterVoting: vote · results ↗ (election 6tthfv, Test ID BV2252).
A live, real-world STAR Voting election: which Goodberry's Frozen Custard flavor is the best? Not a constructed teaching case — an actual poll, minted with no seed ballots, collecting real votes from paper ballots printed at the Cary, NC shop and from anyone who scans the QR code.
Goodberry's sells frozen custard, not ice cream — the title says so on purpose. Its permanent menu board carries a couple dozen flavors, and a rotating daily calendar adds specials on top. That makes it a natural STAR ballot: almost everyone likes several flavors, and a choose-one poll would force each voter to throw away all but one of those opinions.
The ballot¶
Ten flavors from the permanent menu board, spread across its families — custard classic, chocolate, nut, coffee, citrus, mix-in, and the Southern one Goodberry's is known for:
| # | Flavor | Why it's on the ballot |
|---|---|---|
| 1 | Banana Pudding | the Southern signature |
| 2 | Butter Pecan | the nut classic |
| 3 | Chocolate Malt | chocolate, the soda-fountain way |
| 4 | Cookie Dough | the mix-in favorite |
| 5 | Jamocha | the coffee lane |
| 6 | Key Lime | tart, unlike anything else on the list |
| 7 | Mint Chocolate Chip | the perennial top-five finisher |
| 8 | Peanut Butter | the other nut lane, and a second-choice magnet |
| 9 | Salted Caramel | the modern favorite |
| 10 | Sweet Cream | the base custard itself — the purist's vote |
Write-ins are enabled, so a voter whose favorite is Blueberry, Nutella, plain Chocolate, or a calendar special (Lavender Limoncello, Piña Colada, Tiramisu…) can still say so.
How to vote¶
Score every flavor you have an opinion about, 0–5 stars. Give your favorite 5. Give anything you'd be unhappy with 0. Everything in between is fair game — you are rating, not ranking, so equal scores are allowed and normal.
Then STAR does two things:
- Scoring round — add up all the stars; the two highest-scoring flavors become finalists.
- Automatic runoff — every ballot counts once more, for whichever finalist it scored higher. The finalist preferred by more voters wins.
That second step is what a plain star-rating poll lacks: the winner is the flavor more people actually preferred, not merely the one that collected the most generous ratings.
Paper ballots¶
goodberrys_best_flavor_2026_ballots.pdf — 30 print-ready STAR ballots, one per page, each with the 0–5 bubble grid and one big Scan to vote online QR pointing at 6tthfv (the results URL prints as text in the footer). Built PDFs are generated artifacts (*_ballots.pdf is gitignored), so regenerate it from the frozen export with bv_ballot_sheet.py:
.venv/bin/python STARVote_LH_tabulation_engine/tools_adam/bv_ballot_sheet.py --bv-export 01_STAR/04_Real_Elections/goodberrys_best_flavor/cases/goodberrys_best_flavor_2026_bv_export.json --logo STARVote_LH_tabulation_engine/tools_adam/assets/NC_STAR_Logo1.jpg --notice "UNOFFICIAL FLAVOR POLL — COUNTED WITH STAR VOTING." --cover --verify-bv --out 01_STAR/04_Real_Elections/goodberrys_best_flavor/goodberrys_best_flavor_2026_ballots.pdf
--cover prints a preamble page ahead of the 30 ballots. It matters here: anyone who scans the ballot's QR lands on BetterVoting and reads the full election description before voting, so without a cover the paper voter and the phone voter get briefed differently. The cover carries that same text plus how to vote and how it is counted, and says in its footer that it is not a ballot so it can't be hand-counted by mistake. Change --copies for a bigger or smaller table (default 30). Paper and platform stay linked: the QR voters and the bubble-fillers are casting the same ballot into the same count. The full workflow (print → vote → hand-count → compare to BetterVoting) is in running a paper-ballot demo; the by-hand count is in count STAR by hand.
What to watch for when the votes come in¶
- The score leader and the winner may differ. With ten flavors, a polarizing one (Key Lime, Jamocha) can pile up 5s from its fans and 0s from everyone else, lead the scoring round, and then lose the runoff to a broadly-liked flavor. That reversal is the whole point of the "AR" in STAR — see the runoff-reversal case.
- Bullet voting costs you. A ballot that scores one flavor 5 and leaves the rest blank has no say at all in a runoff between two other flavors — see the bullet-vote case.
- Ten candidates, no vote-splitting. Unlike choose-one, adding a tenth flavor similar to the ninth cannot split their support, which is exactly why a long menu is safe on one ballot.
Files¶
| File | What it is |
|---|---|
goodberrys_best_flavor_2026_ballots.pdf |
the printable ballots (30 pages) — generated, not committed |
| cases/goodberrys_best_flavor_2026_bv_export.json | frozen BV export at mint time (zero ballots — the poll had not started) |
Re-freeze the export once real votes are in, and the results are tabulatable in this repo's engine:
uv run STARVote_LH_tabulation_engine/tools_adam/fetch_bv_export.py 6tthfv -o 01_STAR/04_Real_Elections/goodberrys_best_flavor/cases/goodberrys_best_flavor_2026_bv_export.json --force