Taste research for business teams
Ask one question in plain English. In about a minute you get a one-page answer with a verdict, maps and charts. Every number is computed in code and traced to the request that produced it.
How it works
No pickers, no dashboard to configure. Ask the way you would ask a colleague.
Up to 400 characters, read into one of 13 jobs. Names are matched exactly; if one is ambiguous, you pick.
About a minute, step by step: the opening scan, guesses written before any data, follow-ups the data triggers. Stop at any time.
A verdict, rule-scored confidence, key numbers, maps and charts. Every number opens to the request behind it.
Four desks · thirteen jobs
01 · Brands & investing
See which brands really share an audience, how two brands split cities and age groups, and whether a new product idea has a home. Reported SEC figures sit beside the taste data, company-wide.
Not financial advice. No buy, sell or price calls.
02 · Places
Find the neighbourhood whose residents' taste fits your concept, the best tenant for a corner, an area's closest twin, or what sets two to four cities apart. Census and OpenStreetMap counts add who lives there and who is already open.
Shared taste, not foot traffic, rents or sales.
03 · Sponsors & tours
Score sponsors and partners on audience fit, pick launch and tour cities, and find artists and podcasts whose fans look like yours.
Fit only. No fees, pricing or ROI.
04 · Audiences
Who the fans are, how they lean by age and gender, and what else they love, from brands to films to podcasts. Group-level only, never individuals.
No individuals and no sensitive traits.
Also included, no AI model involved
The figures
Each chart states its limit. Each value opens to its request. Missing data is named, never drawn as zero.
Honest by construction
An answer is only worth acting on if you can see where it is thin.
Every figure is computed in code and traced to a logged request. A checker reads the page and states that every number was checked, or says which were not.
It is scored by fixed rules, not by how sure the model sounds. You see what lowered it and the caps it hit. It is not the chance of being right.
No zeros, no flat lines, no guesses filled in. A gap is named in words with the reason.
Each answer shows what a model alone would have said, and the analyst's guesses written before the data, scored after.
Shown apart from the measured answer
Background the model already knows about the brand or place, written in words and kept out of the numbers.
Each finding names its publisher, title and date. It never changes the measured answer.
Rates the data behind this answer, not the chance it is right. The page lists what lowered it.
Limits and access
Every desk, every figure, every source.
Paid plans with higher limits are not open yet. When they are, this page will say so first.
Atlanta, Austin, Boston, Chicago, Dallas, Denver, Houston, Los Angeles, Miami, Nashville, New York City, San Francisco, Seattle
Questions
For analysts: every formula and rule is published.
Read the methodGroup-level taste data from Qloo, US Census ACS, SEC EDGAR filings and OpenStreetMap place counts. Web findings, when shown, carry their publisher and dates and sit apart from the measured answer.
One plain-English question. A one-page answer in about a minute, with every number traced.