Site selection for independent restaurants New York City

Find the block before you sign the lease.

An agent that scouts vacant storefronts and sizes up every rival with Qloo's record of real places: when they close on Friday, what their reviewers rave about, where their fans also go. City inspection records rank the leases.

  • 29 of 30 closing times right with Qloo, 0 of 30 for the model alone
  • Briefs written with Qloo preferred 78% of the time; lease score 0.893 with Qloo, 0.851 without
  • Qloo finds 92.4% of competitors; price levels right for 25 of 27
Scout a location Score any block on the map
26,043 active restaurants 6,767 vacant storefronts

Click any block to score it for a restaurant. Every dot is a real record from New York City.

A lease is the biggest bet a restaurant makes.

Most independent restaurants pick a block by walking past it, then sign a lease that runs for years. Whitespace gives them the read a chain's real-estate team has: the competition, the foot traffic, when every rival closes, and a positioning brief written from what that block's customers like.

  • 3,034restaurants opened in 2025 in the four boroughs and are still operating
  • 6,767storefronts reported vacant to the city, each one a lease someone has to fill
  • 26,043active restaurants, every one a possible rival the score accounts for

What it could earn in New York alone

10%
$299
5%
$50

$0a year

The counts come from New York City records. The shares and prices are assumptions for you to move.

Why Qloo makes the report worth paying for

Scout a location

Describe the restaurant and the people you want to serve. The agent picks neighbourhoods from census data, searches vacant storefronts, ranks every block on city records, sizes up each rival with Qloo, and writes the positioning brief for the top pick.

  • About a minute from plan to shortlist
  • Every step shown as it happens, with every Qloo call
  • Brooklyn, Queens, Manhattan and the Bronx

The agent picks neighbourhoods from census data on who lives and works there, searches vacant storefronts in several, checks competition, and shortlists.

Takes about a minute. Ctrl or Cmd and Enter also runs.

See a finished run

Six real runs from founders' plans. Each one opens in seconds, with every step the agent took.

Scouting

Lease ranking

    The shortlist appears here as soon as the agent has scored the storefronts.

    Drop a pin anywhere in New York.

    Pick a cuisine, then click any block in Brooklyn, Queens, Manhattan or the Bronx. You get the same city-records lease score the agent uses, for that exact block. Drop up to three pins and they compete. No model and no Qloo call: this runs on city inspection records alone.

    Click the map to drop your first pin. Try a busy avenue, then a quiet side street, and watch the score move. Changing the cuisine re-scores every pin.

    What if the block changed?

    Move the sliders to see how the lease score responds to a busier block or more competition. They start from your last pin.

    20
    3
    -of NYC restaurant blocks score lower for this cuisine.

    The agent knows real places because Qloo does.

    A language model can reason about a neighbourhood. It cannot know when the place down the block closes tonight. We asked two models when 30 real New York restaurants close on Friday, alone and with Qloo's record, and checked every answer against Google Maps.

    0 of 30 right for GPT-6 Astra alone29 of 30 right with Qloo

    This is why Whitespace answers every hours question from Qloo and never from the model's memory.

    See all 30 answers

    Loading the recorded answers…

    The duel, restaurant by restaurant

    Pick one of the 30 restaurants from the closing-time check and watch four agents answer the same question.

    The answers recorded in the pre-registered test; nothing is sent to a model.

    Two games, played on real results.

    First, guess when real New York restaurants close on Friday against GPT-6 Astra alone and with Qloo's record. Then pick who scored higher, Qloo or the baseline, on the tests behind the product: the lease score, the briefs and the closing times.

    Beat the model

    Round 1 of 5

    Loading the restaurants…

    10:00 pm

    Qloo or the baseline?

    Question 1 of 7
    You 0/0

    Loading the results…

    Qloo against the baseline, test by test.

    Every test was written down before its result. Red is Qloo, blue is the baseline it was measured against. Hover any chart for the detail.

    Every head-to-head: where the baseline stood, and where Qloo took it One row per comparison, sorted by the size of Qloo's lead. Scores run from 0 to 1 (AUC, accuracy, preference share, share of possible gain, correlation); the metric is named on each row.
    Baseline Qloo, or a model with Qloo
    The full record: results where Qloo did not move the number

    What the agent does

    Six steps, in this order, on every run. Each one is shown on the page as it happens.

    1. Match the concept

      Your concept is matched to one of 25 cuisines the lease score was tested on. A cuisine we did not test stops with a plain message.

    2. Find storefronts

      Describe your plan and the agent picks neighbourhoods from census data, then searches vacant storefronts in several. Or search one area, or bring two to five addresses of your own.

    3. Rank on city records

      Each storefront is scored by how busy its block is and how much of the same cuisine is in and around it, from NYC inspection records.

    4. Write the brief from Qloo, then check it

      A positioning brief for the top pick from what Qloo knows about the area, against a New York baseline. Every business it names is traced to Qloo, to city records near the address, or marked unverified.

    5. Size up every rival with Qloo

      When each nearby competitor closes on Friday and what its reviewers mention, from Qloo's place records, on a closing-time timeline. Plus partners and next steps from city data.

    6. Answer follow-ups

      Ask anything about the shortlist. Hours questions are answered from Qloo, never from the model's memory, and numbers that appear in no tool result are flagged.

    What each source does

    Each source is used only for the job a test says it does well.

    City records

    Where to open

    NYC Health Department inspection records and the Storefront Registry of vacant storefronts. With Qloo the lease score reaches AUC 0.893, against 0.851 without it.

    Census and jobs data

    Who lives and works nearby

    Target customers come from the census. Only 2 of 7 Qloo audience tags matched it.

    City records

    Who the competitors are

    NYC inspection records list every rival; Qloo finds 92.4% of them too, with their hours and reviews.

    Qloo place records

    When competitors close, and what reviewers mention

    City records have none of this. A person checked the closing times on Google Maps: 29 of 30 matched. Price levels matched for 25 of 27.

    Qloo affinity, not validated

    Where else a restaurant's fans go

    Shown as Qloo's list and labelled as not validated. It is never used to choose where to open.

    Where else do their fans go?

    Name up to three places your customers already love. Qloo returns the places their fans also go, and the music and brands that crowd follows. No model is involved.

    How we know

    Every test was written down and committed before its code or its result. Here is what they show, and the full record is at the bottom.

    Which signal found the blocks where each cuisine opened next, 2024 to 2026 AUC within each cuisine, all restaurant blocks in four boroughs. 0.5 is a coin flip, 1 is perfect.
    City-records lease score Qloo cuisine taste

    Where Qloo is strong

    Facts about real places: Friday closing times right 29 of 30, price levels 25 of 27, 92.4% of competitors found. Asked about real restaurants, a model with Qloo's record was right 29 times out of 30, and alone at most 3. Briefs written with Qloo were preferred 78% and 76% of the time.

    How that becomes the product

    The lease score ranks every storefront and reaches 0.893 with Qloo, against 0.851 without it. Every hours question is answered from Qloo's record, never from the model's memory. The brief for the top pick is written from Qloo, and every business it names is traced to Qloo or city records.

    What we kept honest

    Who lives in an area comes from the census, because only 2 of 7 Qloo audience tags matched it. We make no claim about which restaurants survive. The tests where Qloo did not move the number stay in the record below.

    Every test in order, 1 to 28