Skip to content

The resemblance map & neighborhood detail

When you select a brand, LookAlike displays a national census-tract choropleth — a map where neighborhoods are colored by how well they match the brand’s pattern.

The map uses a diverging color ramp. Only neighborhoods in the top fit range are colored:

  • 50–68th percentile: nearly transparent (dark background shows through). These neighborhoods exist but do not visually draw attention.
  • 68–80th: gold/tan; starting to light up.
  • 80–88th: orange; increasingly high fit.
  • 88–95th: red-orange; strong fit.
  • 95–100th: deep red; very high fit to the brand’s pattern.

Why this design: The map intentionally leaves the median tract dark. It highlights where to look, not the entire country. By color, neighborhoods at the 50th percentile are transparent; only the top tier of matches become visible.

Every colored tract shows a within-brand percentile rank (0–100). This number means:

  • 0–49th percentile: Below the national median for this brand. Uncolored on the map.
  • 50–79th percentile: Around or slightly above median. Visible but dim.
  • 80th percentile: High fit; strong signal.
  • 90th+ percentile: Exceptional match to the brand’s pattern; the most promising discovery.

If you toggle Show stores in the map controls, the brand’s actual store locations appear as white circles with a colored ring (the color matches the brand). This lets you:

  • See where the brand already operates (which tracts have stores).
  • Visually cross-reference store locations with the fit ranking.
  • Spot geographic gaps: high-fit regions with no stores yet.

Stores are placed on the map from the brand’s store source and represent the store locations used to fit the brand’s presence model.

Click any tract on the map to open the detail panel on the right. It contains nine sections, top to bottom:

1. Header: tract name, percentile, and status

Section titled “1. Header: tract name, percentile, and status”

At the top you see:

  • Tract label (a Census-named area, e.g., “Greenville-Spartanburg, SC”).
  • Fit percentile in the brand’s color (e.g., “82nd percentile”). At or above the 99th percentile it is shown to one decimal place, so the top of the list does not collapse into a wall of identical “100th” readouts.
  • A fit band chip — Strong, Moderate, or Weak — next to the percentile. It compares the tract’s score with the brand’s own store tracts. Its scale is specific to that brand; it is not an absolute measure shared across brands. See how it works.
  • A small ⓘ glyph next to the percentile. Hovering it shows a one-line caveat: the number is resemblance to the brand’s store neighborhoods on open Census features, not a prediction of sales, foot traffic, or success, and it points to the Method & data page for the full explanation.
  • “Resemblance to [Brand]‘s store neighborhoods” — a reminder of what the score measures.
  • Status tags:
    • “Already has a [Brand]” (if the selected tract contains an existing store).
    • “No [Brand] in this tract” (if open for potential expansion).
    • Distance to the nearest brand store (e.g., “nearest: 2.4 km”).
    • A low ACS reliability badge when the tract has a small population, a missing income estimate, or a large income margin of error — a signal to treat its feature values as noisy.
  • A “Not a fit?” control to flag a lookalike that is wrong on the ground; flagging hides it from the ranked list, and once you have flagged a few, the tool surfaces the features your rejected tracts most share.
  • A star to save the neighborhood to your shortlist.

2. Winners vs. failures (custom upload only)

Section titled “2. Winners vs. failures (custom upload only)”

If you’ve uploaded your own store data with a performance or status column, this section appears. It shows two bars:

  • Resembles your winners (green bar): the percentile of how well this tract fits the model of your high-performing stores.
  • Resembles your failures (red bar): the percentile of how closely it matches stores you flagged as failed or low-performing.

A neighborhood with a high “winners” score and low “failures” score is a strong prospect under your own ground truth.

3. Neighborhood fingerprint vs. the typical store tract

Section titled “3. Neighborhood fingerprint vs. the typical store tract”

A radar (spider) chart compares the tract against the brand’s typical store tract across the 18 scored features:

  • Blue/colored area = the brand’s typical store tract (the median of the tracts where the brand has stores).
  • Light gray area = this specific tract’s feature values.

The 18 scored features are displayed as axes: Income, Education, Density, Age & household, and Retail Context (grocery, restaurants, cafés, shopping). Where the two polygons overlap closely, the tract matches the brand. Where they diverge, you see the differences.

The radar is most useful for visual pattern matching: you can see at a glance whether the tract is denser/older/wealthier than the brand average, or matches it across the board.

A small colored chip labeled:

  • Robust (green): This neighborhood’s high rank is stable across different weighting scenarios. It matches the brand on fundamental characteristics.
  • Fairly robust (yellow): The match holds up under most weighting adjustments, but is somewhat sensitive to theme priorities.
  • Weighting-sensitive (orange/red): This neighborhood’s high rank depends heavily on how certain themes (e.g., Affluence, Urbanism) are weighted. It may drop if priorities change.

The robustness test re-scores the tract under equal weighting of themes and with each theme removed one at a time. If it stays in the high percentiles across all variations, it is robust.

Shows which of the tracked curated brands have a store within 6 kilometers of this tract’s center.

Each entry shows the brand, the distance, and a colored dot. This is a retail-setting snapshot: it shows how much chain retail surrounds the tract, in every category. Only the brands chosen as competitors under Filters count for the competitor filter and the candidate rows.

If the list is empty, no curated brand has a store within 6 km.

A list of the features that raise the score most — the largest positive contributions. Each feature’s contribution is its linear term plus its squared term, measured against an average US tract.

For each feature, you see:

  • The feature name (e.g., “Median Household Income”).
  • The tract’s value (e.g., “$67,400”).
  • A visual bar showing the size of the contribution.
  • A plain-English comparison with the brand’s typical value: “Close to [Brand]‘s typical $65,000” or “Higher than [Brand]‘s typical.”

The features that lower the score most — the most negative contributions:

  • Feature name and the tract’s actual value.
  • A visual bar showing the size of the contribution.
  • A comparison with the brand’s typical value: e.g., “Lower than Chipotle’s typical 22% renters” or “Higher than Trader Joe’s typical median age.”

Use this section to spot potential friction. If a brand typically operates in dense, walkable neighborhoods and this tract has a very low transit-commute percentage, that is a material difference worth investigating further.

8. Full feature comparison (by theme group)

Section titled “8. Full feature comparison (by theme group)”

A table organized by the 5 scored themes:

  1. Income, wealth & housing cost: median household income, per-capita income, median home value, median gross rent, poverty rate.
  2. Education: bachelor’s degree or higher percentage.
  3. Density & transportation: population density, households without a vehicle, walk/bike/transit commute, renter-occupied housing.
  4. Age & household type: median age, average household size, households with children.
  5. Retail & dining context: all POI density, grocery stores, restaurants, cafés, shopping/apparel (per km²).

Each row shows:

  • Feature name.
  • This tract value with an inline bar (gray) showing where it falls on the national 1st–99th percentile.
  • [Brand] typical value with an inline bar (brand color) showing the brand’s typical store-tract value.

This table is the most detailed comparison; use it when you need to audit a neighborhood thoroughly or explain results to a team.

A short paragraph summarizing the neighborhood:

  • For a strong match (82nd+ percentile): “On observable Census features, this neighborhood closely resembles [Brand]‘s store neighborhoods, especially on [top 3 features]. It diverges most on [biggest gap]; if you are considering this tract for other reasons, that is the first thing to check.”
  • For a modest match (below 82nd percentile): “This neighborhood is only a moderate resemblance overall ([X]th percentile). Its closest features to [Brand]‘s are [top 3]. It diverges most on [biggest gap].”

The assessment always concludes with: “This is a description of resemblance, not a prediction that a store would perform here.”

The detail panel no longer repeats a paragraph-long disclaimer at the bottom. Instead the same caveat is attached to the ⓘ glyph on the score line (section 1) as a hover tooltip, and the full explanation lives one click away on the Method & data page, opened from the link at the bottom of the left rail. The message is unchanged: a high percentile means the neighborhood looks like places the brand already operates; it does not predict sales, foot traffic, or that a store would succeed here. Observable neighborhood fit is a weak predictor of revenue, so the tool is for discovery and triage, not site selection alone.

How the map updates with filters and weights

Section titled “How the map updates with filters and weights”

If you adjust the factor weight sliders (Balanced, Affluence-led, Urbanism-led, Household-led), the map re-scores live:

  • Colors shift as themes are prioritized differently.
  • Percentile ranks change for each tract.
  • The detail panel updates to reflect new drivers and gaps.

This is a scenario analysis tool, not a claim about causal importance. Dragging the Urbanism slider up to 2× does not mean urbanism is twice as important in reality; it means “let me see what neighborhoods would rank highest if I cared more about density and transit.”

Similarly, if you compare two brands, the map becomes a diverging map where:

  • Hue indicates which brand a tract leans toward, using a fixed, colorblind-safe blue-and-orange pair (blue for brand A, orange for brand B) rather than the brands’ own colors, so a red-versus-green pairing can’t hide the lean direction.
  • Brightness shows how strongly the tract resembles either brand.

A neighborhood that is very bright and strongly blue is an excellent match for brand A’s pattern. A dimly colored neighborhood is a weak match for both.

  • Fit percentiles are not comparable across brands. Trader Joe’s 85th percentile is not equivalent to Starbucks’ 85th percentile; each brand has its own distribution.
  • The scoring is observable fit only. The tool does not account for foot traffic, local business conditions, permitting, labor costs, or anything not captured in Census and POI data.
  • ACS Census data carries a 5-year lag and sampling error. Smaller tracts have wider confidence intervals. Always pair results with current local knowledge.
  • Store coverage is good for large chains but imperfect. A few stores may be missing or mislocated, which can shift the fit slightly.
  • High fit does not predict sales. See the scope rule in what it is.

Once you have identified promising neighborhoods on the map:

  • Star them into your shortlist (use the star icon in the detail panel) for later review.
  • Export a PDF report with per-tract statistics, drivers, and your notes.
  • Upload your own stores if you have performance or failure data; LookAlike will re-score based on your ground truth.
  • Filter by metro area to focus on a specific region and see a ranked list of that metro’s neighborhoods.

See uploading custom data and the method for more detail.