Skip to content

Robustness & nearby chains

When you click a neighborhood on the map, the detail panel shows two additional signals below the fingerprint visualization: Robustness and Nearby chains. Both help you read neighborhood matches more carefully by revealing what drives or complicates a high score.

Robustness: Is the match weight-dependent?

Section titled “Robustness: Is the match weight-dependent?”

Robustness tells you whether a neighborhood’s high fit rank is stable across different weighting schemes, or whether it hinges on one particular theme.

The tool re-scores the tract under two conditions:

  1. Equal weighting: All 5 scored themes count equally (no tuning via the factor sliders).
  2. Omit-one tests: Each of the 5 themes is removed one at a time while the others use default weighting.

Robustness compares these baseline scores against the neighborhood’s current rank (under your chosen weights) and assigns a label:

  • Robust (≥ 85th percentile under all conditions): The neighborhood scores high no matter how you weight the themes. It genuinely matches the brand’s pattern across multiple dimensions.
  • Fairly robust (≥ 70th percentile under most conditions, but drops below 70 when one theme is removed): The match holds up across most weightings, but has one theme it depends on. The label tells you which one — for example, “weakest when Density & transportation is dropped (falls to 64th).” If that theme matters to your site selection, this is a good signal; if it does not, look elsewhere.
  • Weighting-sensitive (< 70th under all conditions): The neighborhood’s high rank is fragile and leans heavily on a single theme. The label reads: “its high rank leans on [theme] — drop that theme and it falls to [percentile]th.” This is useful context for risk assessment: the match depends on assumptions that may not hold in your analysis.

Robustness is a measure of internal consistency, not external validity. A tract can be very robust in fit to Whole Foods’ pattern but still be a poor site for reasons robustness cannot capture — regulatory environment, real estate availability, competitive saturation outside the 6-km window. Robustness is useful to filter noise from your scoring and to understand what drives each match. Use it alongside the full feature comparison table to decide whether the theme it hinges on matters to you.

Nearby chains: Which tracked brands are close?

Section titled “Nearby chains: Which tracked brands are close?”

Nearby chains shows you competitive density by revealing which of the tracked curated brands have a store within 6 kilometers.

For each tracked brand, the tool finds the nearest store in that brand’s store dataset and calculates the distance to the selected tract’s centroid. Brands within 6 km are listed with their distance rounded to the nearest 0.1 km (or meters if closer than 1 km). If no tracked brands are nearby, the panel reads: “None of the tracked brands have a store within 6 km — a relatively uncontested pocket (among these brands).”

What it tells you:

  • Competitive density among tracked brands: A neighborhood with 4 or 5 of the tracked brands nearby is more saturated than one with none, at least among this sample.
  • Whether the brand already competes locally: If you are considering Trader Joe’s and three other tracked chains are within 6 km, the area is already retail-dense and possibly competitive for this type of shopper.

What it does NOT tell you:

  • Total competitive threat: the tracked curated brands are just a sample; there may be unlisted competitors (regional chains, independents, mass-market grocers) that are closer or more direct threats.
  • Store viability: A sparse competitive zone (no tracked brands nearby) does not mean a store would succeed — it only means those particular brands are not proximate. The area could be rural, low-traffic, or underserved for other reasons.
  • Cannibalization risk: Proximity is not the only driver of cannibalization. A Costco 5 km away may compete for a different shopper profile than a Whole Foods. Use compare mode to see which brand’s pattern the neighborhood matches more closely.

The nearby-chains readout above describes one selected tract. The same idea is also available as two filters on the whole candidate list, so you can screen for the two most routine triage questions directly:

  • At least N km from your own stores. A slider keeps only candidate tracts whose nearest store of the active brand is at least the chosen distance away. Set it above zero to hide neighborhoods so close to an existing location that a new store would cannibalize it. (When no store points are loaded for the brand, the filter keeps the tract rather than guessing.)
  • No competitor within 6 km. A toggle keeps only tracts where no competitor has a store within 6 km. Competitors are the other curated brands in the active brand’s category, such as the other 18 grocers for Trader Joe’s. Select Choose competitors under the toggle to add or remove brands, for example to count warehouse clubs as grocery competitors. Other chains, such as coffee shops or banks, do not count. For your own store list, choose the competitors first.

Each candidate row also carries the distance to the nearest own store and the names of any competitors within 6 km, and both are included as columns in the CSV export, so you can sort or filter further in a spreadsheet. A copied map link keeps a custom competitor choice.

Using robustness and nearby chains together

Section titled “Using robustness and nearby chains together”

A neighborhood that is both robust (independent of weighting) and has no nearby competitors among the tracked brands is a strong triage signal: it resembles the brand across multiple features and sits in a less-contested space. By contrast, a weighting-sensitive match in a densely competitive zone is lower-confidence for site selection purposes, though it might still be worth investigating if the theme it hinges on is central to your strategy.

Remember: both signals are descriptive, not predictive. Resemblance is a necessary condition for site viability, not a sufficient one.