What is site selection analysis?
Site selection analysis is the systematic process of evaluating potential locations for a new facility — a retail store, warehouse, clinic, or any place-based investment — using spatial data and quantitative models rather than intuition alone.
The goal is to identify locations where demand, accessibility, competitive dynamics, and operational constraints align to maximize the probability of commercial success.
Why spatial analysis matters for site decisions
Location decisions carry outsized risk. A retail site locked into a 10-year lease at $500,000 per year represents a $5 million commitment. Location quality is widely recognized in retail geography literature as a primary driver of store-level performance, alongside format, pricing, and assortment (see Birkin et al., Retail Geography and Intelligent Network Planning, Wiley).
Traditional approaches — driving the market, relying on broker recommendations, or using simple radius counts — leave substantial value on the table because they cannot systematically account for:
- Network effects between existing and proposed locations (cannibalization)
- Drive-time polygons that reflect actual accessibility rather than straight-line distance
- Demand heterogeneity at sub-census-tract granularity
- Competitive saturation measured as supply-to-demand ratios, not just competitor counts
The six-phase methodology
Phase 1: Define success criteria
Before any spatial analysis, establish what a “good” location means for this specific use case. Common criteria include:
| Criterion | Metric example |
|---|---|
| Market demand | Households within 10-min drive with income > $75k |
| Accessibility | Average drive time from population centroid |
| Visibility | Traffic count on adjacent arterial (from DOT data) |
| Competition | Ratio of demand to existing supply within trade area |
| Co-tenancy | Presence of complementary businesses (anchor effect) |
| Operational | Zoning compliance, lot size, ceiling height |
The weighting of these criteria varies significantly by industry. A quick-service restaurant chain prioritizes drive-through accessibility and daytime population. A medical clinic prioritizes residential density within a 15-minute drive. A logistics hub prioritizes proximity to interstate interchanges and labor availability.
Phase 2: Delineate trade areas
A trade area defines the geographic extent from which a location draws its customers. The most common approaches:
Ring analysis (simple but imprecise): Fixed-radius circles (1 mi, 3 mi, 5 mi). Ignores road networks, natural barriers, and travel behavior.
Drive-time isochrones: Polygons representing all reachable points within N minutes of drive time, computed from the road network using routing engines (OSRM, HERE, Google). Accounts for highway access, one-way streets, and average speeds.
Gravity models (Huff model): Allocate demand probabilistically based on distance decay and attraction factors (store size, brand strength). Each consumer has a probability of visiting each store proportional to its attractiveness divided by its distance (or travel time), raised to a decay parameter.
The Huff model formula:
P(consumer i visits store j) = (A_j × D_ij^(-β)) / Σ(A_k × D_ik^(-β))
Where A is attraction, D is distance/time, and β is the distance-decay parameter calibrated from observed behavior.
Customer-derived trade areas: When transaction data with customer addresses is available, actual trade area boundaries can be derived empirically. This is the gold standard but requires operational data from existing locations.
Phase 3: Assemble spatial datasets
Effective site selection integrates data from multiple sources:
| Data type | Common sources |
|---|---|
| Demographics | Census Bureau (ACS 5-year), Esri Updated Demographics, CACI |
| Consumer spending | Bureau of Labor Statistics CEX, credit card panels |
| Traffic counts | State DOT, StreetLight Data, Replica |
| Points of interest | OpenStreetMap, SafeGraph (now Dewey), Overture Maps |
| Parcel boundaries | County assessor records, Regrid, CoreLogic |
| Zoning | Municipal GIS open data portals |
| Competition | Manual audit, brand websites, franchise disclosure documents |
All datasets must be geocoded to a common coordinate reference system and aligned temporally. Mixing 2019 demographics with 2026 traffic counts introduces systematic bias.
Phase 4: Score candidate locations
Multi-criteria evaluation (MCE) produces a composite score for each candidate site. The standard approach:
- Normalize each criterion to a 0–1 scale (min-max or z-score)
- Weight criteria according to Phase 1 priorities (weights sum to 1.0)
- Combine using weighted linear combination: Score = Σ(w_i × v_i)
- Rank candidates by composite score
More sophisticated approaches use:
- Analytic Hierarchy Process (AHP) for pairwise comparison of criteria importance
- Boolean overlay for hard constraints (e.g., must be zoned commercial, must have > 5,000 SF)
- Machine learning (Random Forest, XGBoost) trained on performance data from existing locations to predict revenue at new sites
Phase 5: Validate with network analysis
Individual site scores must be evaluated in the context of the existing network. A site that scores highly in isolation may cannibalize an adjacent existing location.
Network-level considerations:
- Cannibalization modeling: Estimate revenue transfer from existing locations to the proposed new site using the Huff model applied to the full network
- Incremental revenue: Total network revenue after opening minus total network revenue before
- Coverage optimization: Maximize population covered within service-level targets while minimizing overlap
- Scenario modeling: Evaluate multiple opening sequences and combinations
Phase 6: Sensitivity analysis and recommendation
Before final recommendation, test how sensitive the rankings are to:
- Changes in criteria weights (±20%)
- Different distance-decay parameters
- Exclusion of uncertain data sources
- Competitive entry scenarios (what if a competitor opens nearby?)
A robust recommendation is one where the top-ranked site remains top-ranked (or in the top tier) across a range of plausible assumptions. If rankings flip under small perturbations, the decision requires additional data or field validation.
Common pitfalls
Over-reliance on demographics alone. Population density and income predict demand but not accessibility or competitive dynamics. A dense, wealthy area may already be saturated.
Ignoring temporal patterns. Daytime vs. nighttime population differs dramatically in mixed-use areas. A breakfast café needs residential density; a lunch spot needs office workers.
Using straight-line distance. In any geography with rivers, highways, or mountains, Euclidean distance severely misrepresents actual travel behavior. Always use network-based travel time.
Confusing correlation with causation in analog models. The fact that your best-performing stores are near highways doesn’t mean highway proximity causes success — it may correlate with population growth patterns in your current markets.
Tools and technology
Modern site selection analysis typically uses a combination of:
- GIS platforms: QGIS (open source), ArcGIS Pro, or cloud-native tools for spatial data management and visualization
- Routing engines: OSRM, Valhalla, or HERE API for travel-time computation
- Statistical software: Python (scikit-learn, geopandas) or R (sf, tidymodels) for modeling
- Visualization: Kepler.gl, Mapbox, or Leaflet for interactive stakeholder presentations
- Data warehouses: PostGIS, BigQuery GIS, or Snowflake Geospatial for large-scale data integration
When to engage a specialist
Organizations benefit most from external geospatial expertise when:
- Expanding into unfamiliar markets where local knowledge is limited
- Opening 5+ locations simultaneously (network effects become complex)
- The capital at risk per location exceeds $1 million
- Internal GIS capability is limited to mapping, not spatial modeling
- Competitive dynamics require sophisticated game-theoretic modeling
References
- Birkin, M., Clarke, G., & Clarke, M. Retail Geography and Intelligent Network Planning. Wiley, 2002.
- Church, R. & Murray, A. Business Site Selection, Location Analysis, and GIS. Wiley, 2009.
- Huff, D.L. “Defining and Estimating a Trade Area.” Journal of Marketing 28(3), 1964, pp. 34–38. doi:10.2307/1249154
- US Census Bureau American Community Survey — Primary source for demographic trade-area data.
- Esri Location Intelligence — Platform documentation for GIS-based site selection.