Dessert Match-ups: Part II
Food
Sweets on sweets on sweets
How decisive were these results, combined?
TL;DR: Ice Cream Sandwich buried Popsicle (36–11, 77%), the most decisive result of this dessert round. Cinnamon, meanwhile, ended in a flat-out tie: Churro 23, Cinnamon Bun 23 — you can’t get closer to a coin flip than that.
| question | N | Leader (votes) | Trailer (votes) | Leader share | Margin (votes) | Margin (points) | p vs 50/50 | Leader 95% CI | Votes to flip |
|---|---|---|---|---|---|---|---|---|---|
| Chilled and Molded | 43 | Flan (30) | Panna Cotta (13) | 69.8% | 17 | 39.5% | 0.013700 | [0.54, 0.83] | 9 |
| Cinnamon | 46 | Churro (23) | Cinnamon Bun (23) | 50.0% | 0 | 0.0% | 1.000000 | [0.35, 0.65] | 1 |
| Cookie | 42 | Macaroon (23) | Macaron (19) | 54.8% | 4 | 9.5% | 0.644000 | [0.39, 0.70] | 3 |
| Summer Treat | 47 | Ice Cream Sandwich (36) | Popsicle (11) | 76.6% | 25 | 53.2% | 0.000346 | [0.62, 0.88] | 13 |
| Tart/Torte | 47 | Fruit Tart (26) | Chocolate Torte (21) | 55.3% | 5 | 10.6% | 0.560000 | [0.40, 0.70] | 3 |
Does HQ agree with RTP?
| question | Leader (overall) | HQ share | RTP share | Gap (pts) | p-value |
|---|---|---|---|---|---|
| Chilled and Molded | Flan | 76% (22/29) | 57% (8/14) | 19% | 0.369 |
| Cinnamon | Churro | 52% (16/31) | 47% (7/15) | 5% | 1.000 |
| Cookie | Macaroon | 61% (17/28) | 43% (6/14) | 18% | 0.443 |
| Summer Treat | Ice Cream Sandwich | 74% (23/31) | 81% (13/16) | 7% | 0.859 |
| Tart/Torte | Fruit Tart | 59% (19/32) | 47% (7/15) | 13% | 0.616 |
The Alignment Index itself is unremarkable this week — 0.73, basically flat versus last time, with the cumulative number at 0.81. But look at the flip count instead of the correlation: 3 of this poll’s 5 questions — Cinnamon, Cookie, Tart/Torte — saw HQ and RTP prefer genuinely different desserts. That’s a majority of this week’s questions, the most we’ve seen split by office in a single poll, and it brings the running tally to 7 flips since RTP joined. Still nothing here clears statistical significance on its own (these are small offices, and small offices are noisy) — but a correlation coefficient can look perfectly fine while more than half the individual questions quietly disagree underneath it, which is exactly what’s happening this week.