Retail Site Selection Analytics – Choose High-Performing Locations

Retail expansion decisions still carry outsized risk because the store is not just a sales point—it is a cost structure (rent, staff, utilities, inventory flow) tied to one postcode. Even in markets where online shopping is mature, physical retail remains the larger share of total spending. For example, the U.S. Census Bureau reported that e-commerce accounted for 16.1% of total U.S. retail sales in 2024, implying the majority of sales still happen through non-e-commerce channels. That is one reason location analytics stays relevant: the wrong site can underperform for years, while the right site can create repeatable growth.

A useful way to bring discipline to site selection is the gravity model—a transparent method for estimating how likely customers are to choose one store over another based on “attractiveness” and “distance”.

1) What a gravity model is, and why it fits retail decisions

A gravity model borrows a simple idea from physics: bigger “masses” attract more, and attraction weakens with distance. In retail, “mass” is store attractiveness (size, range, brand pull, parking, anchor tenants), and distance is travel friction (minutes, kilometres, or cost). The Huff Model is a well-known gravity model that estimates the probability of customers visiting a site as a function of distance and attractiveness, while also accounting for competing sites.

The value of this approach is not that it predicts the future perfectly. It is that it gives you a defensible baseline that can be explained to non-technical stakeholders—useful for analysts coming from a business analyst course who need to connect commercial logic with measurable inputs.

2) The practical ingredients: demand points, attractiveness, and distance decay

Gravity models need three building blocks:

Demand points (where customers are):
These can be neighbourhood centroids, postcodes, census tracts, or mobile-derived “home/work” clusters. Each point carries an estimate of potential demand (population, households, category spend, or customer counts).

Attractiveness (why a store is chosen):
Attractiveness is usually a weighted score. For a supermarket it might include store area, assortment depth, fresh produce quality proxy, parking capacity, and whether it sits in a high-traffic retail cluster. For a QSR brand it might include frontage, visibility, and nearby offices/colleges.

Distance decay (how quickly demand drops with travel):
Most people will travel further for a weekly grocery trip than for a coffee. The Huff approach commonly uses a distance-decay function that reduces probability as travel distance or time rises, and it can be calibrated to reflect different shopping behaviours.

A helpful way to explain this in plain English: “Every extra minute of travel reduces the odds of a visit, but the reduction is not the same for every category.”

3) Turning the model into a store-performance forecast

A gravity model becomes decision-grade when you go beyond a heat map and translate probability into expected performance.

Step A: Calibrate using existing stores
Do not guess parameters if you can learn them. If you have historical store sales and a view of local competition, you can fit the model so that predicted market shares resemble observed shares. Research has shown gravity-model variants can be trained on large-scale mobility and spending datasets to predict customer response and even revenue impacts under scenarios such as store closure.

Step B: Convert probability into sales potential
For each demand point, estimate category spend (or a proxy like household count × spend per household). Multiply that by the probability of choosing the candidate site. Sum across points to get a rough “capturable demand”. This is not final sales, but it is a consistent comparator across locations.

Step C: Stress-test with scenarios
Retail reality changes: a competitor opens, a road becomes congested, a mall loses an anchor tenant. Gravity models are well suited for “what if” analysis because you can adjust attractiveness or add/remove competitors and see how predicted share shifts. Trade-area guidance for business districts notes gravity modelling as a way to predict shopping probability while accounting for competition and travel distance.

This scenario discipline is exactly the kind of structured thinking expected in a business analysis course: define assumptions, model impact, and show decision consequences clearly.

4) Where gravity models work well, and where they need help

Strong fit

Cannibalisation planning:If you open Store B near Store A, how much volume shifts internally versus coming from competitors?
Format decisions:Smaller convenience format vs larger destination format; the attractiveness score changes, and so does the catchment.
Network optimisation:Which stores are redundant, and which locations are “coverage gaps”?

Limitations to acknowledge (and how to reduce them)

Attractiveness is partly subjective.Improve it with observable proxies (store size, parking spaces, tenant mix, ratings, footfall).
Distance is not “as the crow flies”.Prefer drive-time or transit-time where possible.
Not all demand is equal.Segment demand points (income bands, family vs student areas, weekday vs weekend patterns) to avoid averaging away reality.

A practical tip: use a gravity model as the interpretable “spine” of the decision, then add refinement (regression, uplift modelling, or local footfall signals) once the baseline is stable and trusted.

Concluding note

Gravity models are not just academic tools; they are a disciplined way to translate location, competition, and shopper behaviour into a comparable forecast across candidate sites. Their strength is transparency: you can explain why Site X beats Site Y, identify what assumptions drive the result, and run scenarios before committing capital. For teams learning structured decision methods through a business analyst course or a business analysis course, this approach is a practical bridge between commercial judgement and measurable, repeatable analytics.

Business Name: Data Analytics Academy
Address: Landmark Tiwari Chai, Unit no. 902, 09th Floor, Ashok Premises, Old Nagardas Rd, Nicolas Wadi Rd, Mogra Village, Gundavali Gaothan, Andheri E, Mumbai, Maharashtra 400069, Phone: 095131 73654, Email: elevatedsda@gmail.com.