Four ways to combine places and people.

Use when the campaign is anchored to a fixed list of physical locations — your stores, dealers, clinics, franchisee territories, sponsorship venues — and you need to know who lives within a defined radius of each one. The output is the audience around your footprint. Avoid when you don't have a venue list yet (you're still prospecting), when the right answer is a list of POIs rather than CBGs (use Method 02), or when proximity inside the radius matters more than the radius itself — Method 01 treats every CBG inside the circle equally, so a venue 200 feet away counts the same as one 0.9 miles away.

Use when you need a list of POIs, not CBGs — co-marketing prospecting, retail-partner identification, sponsorship targeting, local activation outreach. You're asking "which venues sit inside the audience I care about?" Avoid when the campaign is media-buying first (you need CBGs to traffic against, not a partner list), when your audience definition is too broad to filter meaningfully (everything passes the threshold and you get every POI in the market back), or when you need to weight venues by how well they match — Method 02 is binary: a POI is either inside a high-index CBG or it isn't.

Use when the count of venues in a block is itself a planning signal — urban density plays, retail clusters, nightlife corridors, food courts, strip malls, dispensary corridors. A CBG with 8 coffee shops behaves differently than a CBG with 1, and you want that to show up in the score. Avoid in low-density categories where most CBGs have zero or one POI (the density signal collapses), in rural markets (same problem at scale), or when one nearby venue across a CBG boundary should count for more than three venues clustered inside — that's a Method 04 question, not a Method 03 question.

Use when proximity is the planning signal — walkability-driven categories (coffee, convenience, QSR, urgent care), urban markets where CBG edges are arbitrary, premium and drive-time categories where one nearby venue is worth more than three far ones. Use it when a hard radius cutoff feels wrong. Avoid when you need a clean, defensible "in / out" answer for a stakeholder (decay scores are continuous and harder to explain than "within 1 mile"), when your POI footprint is sparse (decay halos overlap with nothing and the score flattens), or when run time matters and you're scoring nationally — Method 04 is the heaviest of the four because every CBG-to-POI pair is calculated.

Start with Method 01 (Geo-Filter then Score). It's the most intuitive — pick a radius around your venues, score the audience inside. It's also what AMAP runs by default in the data tool today, so the output format is familiar to anyone who's seen an AMAP CBG ranking before.

Method 03 is binary-by-block: it counts how many venues fall inside each CBG. Method 04 is continuous: it weights every venue by distance, regardless of CBG boundaries. Use 03 when block-level density is the planning signal (clusters, corridors). Use 04 when proximity is the planning signal (walkability, drive-time).

Yes. Any of the four methods accept a composite audience input — e.g., Hispanic proportion × HHI $75K+ × Recently Moved. The composite is built from the AMAP taxonomy before the POI math runs. Weights between segments are tunable.

The full Foursquare (FSQ) taxonomy — roughly 1,000+ categories down to chain-level (e.g., Pizza Restaurant > Domino's). You can also pass a custom POI list (your store file, dealer locations, sponsorship venues) instead of an FSQ category.

No. One mile is the default. Common overrides: 0.25mi for walkable urban categories, 3–5mi for suburban auto and big-box, 10mi+ for rural healthcare and ag. The radius is a parameter, not a constraint.

CBGs are roughly 10× smaller than ZIPs (~1,500 people vs ~15,000). For POI work that resolution matters — a single ZIP often spans both a high-Hispanic block and a low-Hispanic block on opposite sides of a highway. CBGs surface that. AMAP also returns ZIP, county, and DMA roll-ups when you need them.

Every method returns a CBG list (or POI list, for Method 02) with a segment ID. That ID is pushed to TTD, DV360, Meta, StackAdapt, Amazon, and Goldfish DOOH ready to traffic. No re-keying, no manual upload.

Zero PII. No cookies. No device IDs. Every score is aggregated at CBG level above a population threshold. The POI layer is venue-level (public locations), never visitation data tied to people.

The skill picks based on the brief. If the brief is ambiguous, it returns the recommended method plus the runner-up and asks one clarifying question. You can also override the method explicitly in the prompt — e.g., "use Method 04, decay capped at 3mi."

Method 01 and 02: under 5 seconds for a single DMA. Method 03: 5–10 seconds. Method 04: 10–30 seconds depending on POI count and market size (it's the most computationally heavy because every CBG-to-POI pair is calculated). National runs scale linearly.