Does it find real accounts?
We tested TasteRoute against an answer key: the venues that already carry real brands, taken from each brand's own published list. Every method ranked the same pool of Qloo's best-known venues, and none saw the brand's name.
TASTEROUTE (QLOO FIT)
8
of 47 accounts in its top 30
AN LLM READING NAMES
8
same pool, same brief
QLOO TASTE ONLY
10
affinity without the mass-market check
CHANCE
1.5
expected from a random order
VERDICT
A tie, not a win
We wrote down the target before running the test brands: TasteRoute beats the LLM on recall@30 in 3 of 4 test brands. TasteRoute won 1, tied 1 and lost 1 of the 3 brands we could score. The fourth could not be scored because Qloo knows none of its comparable brands; the app refuses to rank in that case too. Qloo and the LLM found different accounts, so the results page shows both.
By brand
Share of the brand's accounts in each method's top 30. Dev brands were used to fix the method before the test run.
| Brand | Accounts | TasteRoute | Qloo taste only | LLM | Chance |
|---|---|---|---|---|---|
| Kings County Distillery dev A small-batch distillery making bourbon, rye and moonshine | 10 | 20% | 10% | 20% | 3% |
| Wölffer Estate dev A Long Island estate making rosé wines and dry rosé cider | 7 | 0% | 0% | 0% | 3% |
| Steven Graf Wine test An importer of low-intervention wines from small European and American growers | Not scored: 0 of 6 comparables (Gut Oggau, Martha Stoumen, Broc Cellars, Kermit Lynch Wine Merchant, Scribe Winery, Donkey & Goat) are known to Qloo; the minimum is 3. | ||||
| Parlor Coffee test A Brooklyn specialty coffee roaster supplying cafés and restaurants | 4 | 0% | 25% | 25% | 3% |
| St. Agrestis test Non-alcoholic bottled aperitivo cocktails and a booze-free amaro | 36 | 19% | 22% | 19% | 3% |
| Casamara Club test Non-alcoholic botanical sodas inspired by Italian amari | 7 | 14% | 14% | 0% | 3% |
Different accounts
The test brands' accounts in each top 30. The two lists share 3; together they reach 13 of 47 accounts, either alone 8. Ranks are out of each brand's pool of about 900 venues.
ONLY TASTEROUTE FOUND
5
in its top 30, not the LLM's
- June · St. Agrestis
TasteRoute 2nd · LLM 101st
- Vinegar Hill House · St. Agrestis
TasteRoute 3rd · LLM 46th
- Grand Army · St. Agrestis
TasteRoute 9th · LLM 284th
- Glasserie · Casamara Club
TasteRoute 10th · LLM 37th
- Madeline's · St. Agrestis
TasteRoute 22nd · LLM 283rd
ONLY THE LLM FOUND
5
in its top 30, not TasteRoute's
- Locanda Vini & Olii · St. Agrestis
TasteRoute 62nd · LLM 7th
- Montesacro · St. Agrestis
TasteRoute 84th · LLM 12th
- La Vara · St. Agrestis
TasteRoute 127th · LLM 30th
- Petee's Café · Parlor Coffee
TasteRoute 157th · LLM 23rd
- Cafe Colette · St. Agrestis
TasteRoute 349th · LLM 28th
BOTH FOUND
3
in both top 30s
- Lilia · St. Agrestis
TasteRoute 5th · LLM 1st
- Five Leaves · St. Agrestis
TasteRoute 20th · LLM 22nd
- Cecconi's DUMBO · St. Agrestis
TasteRoute 23rd · LLM 2nd
Less-known venues. 39 of the 47 accounts are outside Qloo's 50 best-known venues. In the top 30: TasteRoute 5, Qloo taste only 7, the LLM 4. We chose this measure before the test; the numbers are small.
Noticed after the test, so an observation rather than a result: the LLM's own finds often say what they sell in their name (“Vini & Olii”, “Cafe”). Several of TasteRoute's don't (June, Grand Army), and the LLM ranks those about 100th and 280th.
AN ACCOUNT QLOO FOUND AND THE LLM MISSED
Grand Army
A real St. Agrestis account (non-alcoholic bottled aperitivo cocktails and a booze-free amaro). TasteRoute ranked this bar 9th of 875; the LLM ranked it 284th. Fans of Ghia and Campari over-index there (Qloo).
What we learned
- Qloo is strongest at the top of the list for taste-led brands: for the non-alcoholic aperitivo brand, Qloo's taste score put 7 of 36 accounts in its top 15, the LLM 4.
- The LLM reads names (“Whiskey”, “Wine House”, “Coffee”) and knows famous venues, and ranks accounts higher on average.
- Our mass-market check counts Starbucks and Dunkin', which pushed real coffee accounts down for a coffee roaster. Qloo's taste score alone did better on this test.
- If Qloo doesn't know the brands you resemble (small natural-wine producers here), there is no taste signal to rank with.
- Music taste did not help. Ranking also by the artists your comparables' fans like found 4 test accounts in the top 30, against 8, so it only explains a match and never changes fit. We wrote that rule down before running it.
- Qloo's heatmap did not help either. Ranking also by where your comparables' fans concentrate found 6 test accounts in the top 30, against 8, so it orders the taste route and never changes fit. That rule was also written down first.
Method and limits
Six real brands with public account lists, two used to fix the method and four run once. The method, comparables and target were committed before the test run, and the one change made after the dev run is recorded. Pools are Qloo's 500 best-known Brooklyn venues per venue type; accounts outside them are left out for every method. Samples are small: one account moves a brand's number by several points.
Full tables, every account and the code →