Home/Methodology

Scoring Methodology

A complete description of the SiteScapr v2 formula - how scores are computed, what data powers each variable, and where proxied estimates are used. Transparency is a design principle here, not an afterthought.

Formula v2.0Kolkata Alpha DatasetLast updated: Feb 2026

Formula Overview

The v2 formula decomposes the Location Score into three interpretable sub-scores. Each sub-score is computed from 2–3 normalised sub-indices and combined with a weighted sum.

Location Score

LS = (Demand × 0.40) − (Friction × 0.35) + (Growth × 0.25)

The top-level composite score. Demand and Growth are positive contributors; Friction is a penalty. The weights reflect the relative importance of market opportunity vs. operating headwinds vs. future trajectory.

Explanation: Your final grade. The more opportunity a neighbourhood offers, the higher it scores - and the harder it is to operate there, the more it loses.

Demand Score

DS = (0.30 × income_index) + (0.35 × foot_traffic_index) + (0.35 × population_density_index)

Measures the revenue potential of the location. Foot traffic and population density are weighted equally and slightly higher than income, because volume-driven businesses depend more on throughput than on per-customer value.

Explanation: How many people want what you're selling, and can they afford it? This measures the sheer volume and spending power of people in the area.

Friction Score

FS = (0.40 × adj_competition) + (0.35 × commercial_rent) + (0.25 × accessibility_penalty)

Measures the operating resistance of the location. Adjusted competition is the dominant friction factor. Rent follows, and accessibility is the smallest but still material component.

Explanation: How much will this location cost you - in rent, rivals, and hassle? A high friction score is a red flag. It gets subtracted from your final score.

Adjusted Competition Index

adj_competition = competition_index × (1 − clustering_benefit_factor)

Reduces the raw competition penalty when the business type benefits from co-location with similar businesses (e.g., food courts, retail clusters). The CBF is set per business type and ranges from 0.0 (professional services) to 0.5 (food & beverage).

Explanation: Being surrounded by similar businesses isn't always bad. A café next to other cafés benefits from the crowd they all attract together - so the competition penalty is softened.

Growth Score

GS = (0.50 × area_growth_trend) + (0.30 × vacancy_rate_improvement) + (0.20 × infrastructure_investment_index)

Forward-looking component capturing the medium-term trajectory of the area. Growth trend dominates; vacancy improvement signals market confidence; infrastructure investment is a leading indicator of long-term desirability.

Explanation: Is this neighbourhood on its way up? This rewards areas showing signs of development, fresh investment, and fewer empty shops - a bet on the future.

Clustering Benefit Factor

Some businesses actually do better when their rivals are nearby - because together they become a destination. This factor adjusts the competition penalty to give those businesses a fair deal.

The CBF reduces the raw competition penalty by the proportion that "cluster draw" offsets head-to-head competition. A restaurant in a food court competes less negatively with nearby restaurants than, say, two law firms on the same street.

Business TypeCBFRationale
Tech Office, Medical Clinic, Educational Institute0.00Professional services compete on differentiation; co-location provides no benefit.
Pharmacy, Gym / Fitness Centre0.15Mild clustering benefit from shared foot traffic corridors.
Retail Store, Supermarket, Salon & Beauty0.30Retail clusters attract comparison shoppers, reducing effective competition.
Restaurant, Cafe0.50Food & beverage hubs create destination effects - competition is strongly mitigated by cluster draw.

Variables & Data Sources

These are the ingredients that go into every score. Each one is a real-world measurement - like how busy a street is, or what rent costs - converted into a number between 0 and 1.

Variables marked proxy are estimated from secondary sources and should be treated as indicative. Primary-sourced variables are updated at least annually.

VariableScaleSourceNotes
income_index
0–1OpenCity / Census approximationAverage household income and purchasing power. Normalised from 0–100 raw scale.
foot_traffic_proxyproxy
0–1Transit density + landmark proximity proxyEstimated pedestrian flow derived from metro station proximity, road grade, and commercial density.
population_density_index
0–1Census 2011, KMDA estimatesNormalised residential + daytime population density.
competition_indexproxy
0–1OSM commercial POI densityCount of similar-category businesses in a 500m radius, normalised to dataset max.
commercial_rent_index
0–1PropTiger / 99acres commercial listing dataAverage per sq.ft. commercial rent, normalised. Rent index 1.0 ≈ ₹5,00,000/month for a typical unit.
accessibility_penaltyproxy
0–1Estimated - road width, metro proximity, parking availabilityProxied from road grade and transit coverage. Higher = harder to access. Not from a primary data source.
area_growth_trendproxy
0–1Estimated - KMDA development plans, satellite change detectionProxied from published development zone classifications and new commercial registration density. Updated annually.
vacancy_rate_improvementproxy
0–1Estimated - commercial occupancy surveysProxy for declining vacancy. Derived from listing age on property portals vs. prior year.
infrastructure_investment_indexproxy
0–1Estimated - state budget allocations, KMDA reportsCaptures planned and recent capital investment in roads, metro, utilities. Reviewed semi-annually.

Normalisation

Every neighbourhood is rated on the same 0-to-100 scale so they can be compared fairly - like grading all students out of 100 regardless of how hard each test was. The final score you see is that same idea applied to location quality.

The Kolkata dataset stores all values on a 0–100 integer scale for readability. The scoring engine divides all values by 100 before applying formula weights, so the internal calculation always operates on a 0–1 continuum.

The final Location Score is multiplied by 100 for display, giving a 0–65 practical range(theoretical max is 65.0 when all positive inputs are 1.0 and all friction inputs are 0.0).

In the Compare calculator, inputs should be entered directly on a 0–1 scale. Values above 1.0 are automatically clamped to 1.0.

Limitations & Assumptions

No model is perfect. Think of SiteScapr as a very well-informed starting point - not the final word. Here is what it cannot see:

  • 1All Kolkata neighborhood data is point-in-time (Feb 2026). Rapidly changing areas (New Town, Rajarhat) may deviate from actuals within 6–12 months.
  • 2The formula is a deterministic weighted model, not a machine learning prediction. It cannot account for hyper-local factors (e.g., a specific corner vs. mid-block on the same street).
  • 3Proxy variables (marked above) introduce estimation error. Treat scores as relative rankings within the dataset, not absolute predictions of revenue.
  • 4Budget filtering uses a linear rent approximation. In reality, commercial landlords negotiate non-linearly and lease terms vary significantly.
  • 5The formula does not currently account for seasonality, zoning regulations, or lease availability - all of which materially affect real-world location decisions.

Try the Formula Yourself

Use the Compare calculator to enter your own sub-index values and test custom weight configurations.

Open Compare →