How we test the score
A fit score is only worth something if it agrees with what brands actually do. We gathered 202 real sponsorships, each with a public source, across 9 podcasts, 8 festivals and 7 teams, plus 122 brands in the same categories that did not sponsor them. We wrote the method down and hashed the answer key before the first run.
What we found
The brands that actually sponsor these 24 properties are no better matched to their fans, by Qloo's taste data, than same-category brands that don't. Somm's Fit Score separates them at chance level, and so does Qloo affinity on its own. A language model with no taste data does better, mostly on deals that were widely announced: it is remembering the status quo, not judging fit.
Somm doesn't predict who already signed, and we don't claim it does. It shows which brands a property's fans over-index for, with a receipt for every number. The test says today's sponsorship money mostly isn't following that signal.
This is run 2. Run 1 had a bug that the negative control caught: a market location signal that drowned out each audience. We fixed it and ran again, and say so in the registration. Both runs are in the repository.
Results
AUC is the chance that a brand that really sponsored the property scores higher than one that didn't: 0.5 is a coin flip, 1.0 is perfect. Run on 10/28/2026.
| Method | Mean AUC | Podcasts | Festivals | Teams | Properties won |
|---|---|---|---|---|---|
Somm fit score The production scorer, all five signals. | 0.49 | 0.51 | 0.52 | 0.42 | 1 |
Qloo affinity only One Qloo Insights call ranking the candidates for this audience. | 0.49 | 0.49 | 0.55 | 0.41 | 1 |
Popularity only How popular the brand is overall, ignoring the property. The baseline to beat. | 0.43 | 0.44 | 0.45 | 0.41 | 1 |
Model only The same language model, asked to rank the same brands with no Qloo data. | 0.64 | 0.59 | 0.68 | 0.68 | 18 |
Negative control Affinity computed against a different property's fans. Should be near chance. | 0.48 | 0.55 | 0.50 | 0.38 | 3 |
Gate traps: 10 of 11 caught
- caughtgt-rollingloud-hennessy verdict blocked. Rolling Loud (Orlando) admits all ages, so a property-wide Spirits deal is out; activation would have to sit inside 21+-gated zones. DISCUS Code of Responsible Practices requires 73.8% of the audience 21+.
- caughtgt-rollingloud-draftkings verdict blocked. Rolling Loud (Orlando) admits all ages, so a property-wide Sports betting deal is out; activation would have to sit inside 21+-gated zones. AGA Responsible Marketing Code for Sports Wagering requires 73.6% of the audience 21+.
- caughtgt-katseye-michelob verdict blocked. Age-restricted category with no audience-age evidence from Qloo; Somm will not clear it blind. Beer Institute Advertising and Marketing Code requires 73.6% of the audience 21+.
- caughtgt-katseye-draftkings verdict blocked. Age-restricted category with no audience-age evidence from Qloo; Somm will not clear it blind. AGA Responsible Marketing Code for Sports Wagering requires 73.6% of the audience 21+.
- caughtgt-govball-fanduel verdict consider. Some youth skew (+0.20 affinity, ages 24 and younger). Youth-skewed acts: A$AP Rocky — keep the brand off their slots. AGA Responsible Marketing Code for Sports Wagering requires 73.6% of the audience 21+. Confirm with measured audience composition before signing.
- caughtex-coachella-pepsi verdict blocked. Category exclusivity: Coca-Cola holds Soft drinks & water for Coachella Valley Music and Arts Festival.
- caughtex-stagecoach-wrangler verdict blocked. Category exclusivity: American Eagle holds Apparel for Stagecoach Festival.
- missedex-dodgers-polymarket . could not resolve property or brand in Qloo
- caughtex-rollingloud-nike verdict blocked. Category exclusivity: Under Armour holds Apparel for Rolling Loud (Orlando).
- caughtex-barcelona-adidas verdict blocked. Category exclusivity: Nike holds Apparel for FC Barcelona.
- caughtctl-coachella-cocacola verdict pursue.
Per property
| Property | Matched in Qloo | Somm fit score | Qloo affinity only | Popularity only | Model only | Negative control |
|---|---|---|---|---|---|---|
| Huberman Lab | 7/10 sponsors, 4/5 controls | 0.57 | 0.46 | 0.29 | 0.75 | 0.21 |
| Lex Fridman Podcast | 5/9 sponsors, 4/5 controls | 0.60 | 0.35 | 0.50 | 0.95 | 0.60 |
| The Tim Ferriss Show | 6/8 sponsors, 3/5 controls | 0.42 | 0.56 | 0.28 | 0.94 | 0.67 |
| The Joe Rogan Experience | 7/8 sponsors, 5/5 controls | 0.53 | 0.63 | 0.29 | 0.49 | 0.60 |
| The Bill Simmons Podcast | 6/6 sponsors, 5/5 controls | 0.60 | 0.57 | 0.73 | 0.40 | 0.40 |
| Pardon My Take | 7/8 sponsors, 5/5 controls | 0.44 | 0.49 | 0.40 | 0.49 | 0.63 |
| Call Her Daddy | 6/8 sponsors, 5/5 controls | 0.42 | 0.27 | 0.57 | 0.63 | 0.47 |
| Armchair Expert with Dax Shepard | 6/6 sponsors, 5/5 controls | 0.50 | 0.60 | 0.37 | 0.63 | 0.60 |
| This Week in Startups | 2/7 sponsors, 2/5 controls | 0.50 | 0.50 | 0.50 | 0.00 | 0.75 |
| Coachella Valley Music and Arts Festival | 10/10 sponsors, 5/5 controls | 0.48 | 0.50 | 0.38 | 0.82 | 0.50 |
| Lollapalooza (Chicago) | 11/11 sponsors, 5/5 controls | 0.66 | 0.53 | 0.53 | 0.78 | 0.42 |
| Austin City Limits Music Festival | 11/11 sponsors, 5/5 controls | 0.42 | 0.45 | 0.42 | 0.76 | 0.49 |
| The Governors Ball Music Festival | 11/12 sponsors, 5/5 controls | 0.60 | 0.64 | 0.53 | 0.67 | 0.58 |
| Bonnaroo Music & Arts Festival | 7/10 sponsors, 5/5 controls | 0.57 | 0.80 | 0.43 | 0.89 | 0.46 |
| Stagecoach Festival | 8/9 sponsors, 5/5 controls | 0.63 | 0.60 | 0.42 | 0.35 | 0.53 |
| Primavera Sound Barcelona | 11/11 sponsors, 4/5 controls | 0.47 | 0.48 | 0.36 | 0.89 | 0.48 |
| Rolling Loud (Orlando) | 8/8 sponsors, 5/5 controls | 0.35 | 0.40 | 0.50 | 0.30 | 0.57 |
| Inter Miami CF | 10/10 sponsors, 6/6 controls | 0.44 | 0.37 | 0.50 | 0.57 | 0.30 |
| Golden State Warriors | 5/7 sponsors, 5/5 controls | 0.28 | 0.28 | 0.44 | 0.72 | 0.48 |
| Los Angeles Dodgers | 4/7 sponsors, 3/5 controls | 0.50 | 0.42 | 0.25 | 0.58 | 0.42 |
| Real Madrid CF | 6/6 sponsors, 5/5 controls | 0.33 | 0.37 | 0.37 | 0.43 | 0.30 |
| FC Barcelona | 4/5 sponsors, 4/5 controls | 0.34 | 0.31 | 0.31 | 1.00 | 0.38 |
| McLaren Mastercard Formula 1 Team | 8/11 sponsors, 6/6 controls | 0.54 | 0.56 | 0.60 | 0.63 | 0.50 |
| Wrexham AFC | 2/4 sponsors, 5/5 controls | 0.50 | 0.60 | 0.40 | 0.80 | 0.30 |
What this does not show
- Fit is not the same as who pays. Budgets, agencies and timing decide deals too, so some real sponsors are poor fits and some good fits never signed.
- A control brand being absent from our sources doesn't prove it never sponsored the property. Podcast ads are inserted per listener, so podcast controls are the weakest.
- The model-only baseline can remember real deals from press coverage. A high score there is memory, not audience reasoning.
- Qloo scores are aggregate audience affinities. They say nothing about any individual and can't certify the 21+ share that alcohol and betting codes require, which is why Somm never clears those deals on taste alone.
How the score is built
- Fan affinity, 35%. Where the brand ranks for this audience among the category's brands.
- Love back, 20%. Whether the brand's fans like the property too.
- Who they are, 20%. Similarity of the two audiences' age and gender lean.
- Where they are, 15%. Overlap of the strongest areas in the market.
- Momentum, 10%. The property's popularity trend over the past year.
- Missing parts are left out and the rest re-weighted. With fewer than three parts, Somm shows no score at all.