Real Estate Site Selection: A Data-Driven Guide

Where to place your company’s next location, whether it is a single site or part of a multi-location expansion strategy, is a momentous decision. Choosing a less profitable location than an alternative you could have gone with means years of unearned revenue, and one painful dud can seriously hurt your company’s bottom line.

This guide covers how data-driven site selection works, from process and data to common mistakes, tools and pricing.

In this guide

What data-driven site selection is and who uses it

Some approaches to real estate site selection predict future revenue far better than others. Some companies rely on chance encounters with vacant lots or empty buildings. Others fail to separate neighborhood factors (who lives and works there) from on-site factors such as accessibility, visibility and facility attractiveness. Still others try to use analytics but are unsure how to evaluate multiple variables, and at what radius levels, so they cannot combine many data points into a single score.

Data-driven site selection replaces those habits with a consistent, mathematical evaluation of markets, neighborhoods and sites.

The model-based approach

Ambient Array’s approach to real estate site selection is based on a single, time-tested observation: a site’s profit potential is primarily driven by who lives and works in the area.

Tupll, our model-based approach, evaluates, weights and combines multiple significant area factors to produce a much more predictive revenue score. We usually start with one or more entire markets, then subdivide the top markets into neighborhoods, or “zones,” and evaluate them mathematically.

Most importantly, we base the model on your own historical data: where you currently operate and how successful you have been in those locations. The resulting model is as unique to your firm as DNA is to an individual. We use advanced statistical techniques to isolate the factors that contribute to your success and throw out the rest.

To see the approach applied, read our haircut chain site selection case study and a second case study in which the bigger market was not the better market.

How self-service software tools differ

Most site selection services available today are self-service tools, usually software as a service (SaaS). You log in, type in an address and press “Go.” The system outputs population counts, demographic breakouts and more, often with maps and a CSV export for further analysis in Excel. These tools are convenient, but they make it easy to fall into the mistakes covered below.

Site selection examples by business type

Site selection questions come in five common forms.

Starting a new brand from scratch. With no established customer loyalty, location matters even more. A boutique clothing store might benefit from a nearby café or salon that attracts similar demographics. Urban startups frequently choose high foot traffic locations slightly off main streets, to balance visibility with affordable rent. Salons and fitness studios often select on demographic match instead of foot traffic alone.

Increasing your footprint in an existing market. Some coffee chains use “clustering” strategies, sometimes placing locations within blocks of each other to dominate a market. Fast food chains might target 5 to 10 minute drive times between stores, to maximize convenience without cannibalizing sales. Banks typically expand based on deposit potential, analyzing household incomes, business density and competitor presence.

Deciding which geographic market to enter next. Large general-merchandise retailers analyze metropolitan areas on population growth, income levels and regional economics. Hotel brands examine business travel patterns, tourism trends and convention activity.

Testing the revenue potential of a particular address. Fashion and specialty retailers use pop-up shops to test customer response before signing a long-term lease. Restaurant entrepreneurs use mobile operations to find the neighborhoods that respond best. Online retailers opening physical locations look for zip codes with high online purchase rates in their shipping data.

Ranking areas within a market by revenue potential. Supermarket chains analyze household density, income levels and competitive saturation, and typically require minimum population thresholds within specific drive times. Convenience stores prioritize traffic counts and look for corner lots with easy ingress and egress on the “going home” side of commuter routes.

Prefer to listen? Listen to our podcast episode on real estate site selection on YouTube.

A process-driven approach in four stages

Site selection teams tend to get caught between two extremes. Some, eager to secure a location, rush and skip crucial steps. Others get overwhelmed by the volume of data and bog down in endless analysis, which delays expansion plans. Working through the key stages in order avoids both.

1. Start with the neighborhood

Begin by analyzing neighborhoods instead of jumping straight to specific properties. Once you have identified the areas with the highest potential for your business, focus on individual properties within them.

2. Fine-tune the neighborhood analysis

  • Determine the correct radius levels. The relevant area varies with your business type and local geography.
  • Gather diverse data points. Pull dozens of demographic, psychographic and commercial factors for each area.
  • Combine the data algorithmically. Use machine learning techniques to weigh and integrate the factors into a model trained on your own historical success factors.

3. Work with brokers strategically

Work with real estate brokers to identify available properties, but do it in a way that minimizes time wasted on less profitable areas.

4. Evaluate on-site factors

Once you have identified promising properties, evaluate each one on site:

  • Accessibility (ingress and egress) for customers, suppliers and deliveries
  • Visibility from major thoroughfares
  • Curb appeal and architectural condition
  • Building dimensions, layout and systems
  • Parking availability and convenience
  • Potential for future expansion
  • Compliance with zoning regulations and building codes

Weigh the financial and regulatory side as well: lease terms and operating costs, signage ordinances, environmental requirements, utility costs, property taxes and insurance.

The data that supports site selection

Three kinds of evidence drive a location decision: neighborhood characteristics, location-specific attributes (covered above) and foot traffic patterns.

Demographic information

Demographic data shows who lives, works and shops near a potential location. The U.S. Census and other demographic databases cover age and gender distributions, household income levels, family structures, educational attainment and occupation types. A high-end retailer might seek areas with higher median incomes and education levels, while a family-oriented restaurant might prioritize neighborhoods with a high concentration of households with children.

Psychographic insights

Demographics tell you who consumers are. Psychographic data helps explain why they make certain choices. It covers the values, attitudes, interests and lifestyle choices that drive purchasing decisions. Learning that local consumers put convenience ahead of price, for example, could influence store layout and service offerings.

Consumer spending patterns

Spending data shows not only how much money consumers have but how they choose to allocate it. Credit card transaction data, mobile payment information and economic reports can show which business categories are thriving or struggling in an area. If the data shows high spending on dining out but few upscale restaurants, that might indicate an opportunity for a fine dining establishment.

Area-specific attributes

Crime statistics help assess security needs and operating hours. Employment rates and business growth patterns give context for long-term viability. Transportation infrastructure, including parking and public transit access, affects customer convenience.

Foot traffic and visitation patterns

Neighborhood data and site attributes rest on assumptions about how people might behave. Foot traffic data, which relies heavily on mobile device data, shows who actually visits an area, and it goes beyond simple counting:

  • Temporal patterns: morning rushes, lunchtime peaks and evening surges, by day, week and season
  • Dwell time: how long visitors typically stay in the area
  • Cross-shopping behaviors: which other businesses people visit during the same trip
  • Loyalty metrics: how frequently the same visitors return
  • Competitor analysis: how foot traffic compares with similar businesses in the region

A coffee shop might find that a location near a popular gym sees consistent morning traffic, while a spot near office buildings gets steady afternoon visits. Trends count too: a location with consistent year-over-year growth in foot traffic may be a better long-term investment than one with higher current numbers but a declining trend.

Why no single data type is enough

These data types work together. High foot traffic does not indicate a good location if the visitors do not match your target customer profile, and a site can check every box for demographics and physical attributes and still underperform because of traffic patterns that only foot traffic analysis would reveal.

Eight common site selection mistakes

Most site selection teams make one or more of these mistakes, and each can put the company in a less profitable location.

Mistake 1: Choosing locations based on circumstantial information

Background: The number of factors that contribute to a location’s success can be overwhelming. Not surprisingly, site selection teams take various homespun approaches to narrowing their options down to a handful of candidate locations.

The Big Costly Mistake: Most companies rely too much on chance to identify candidate locations. Many start by randomly driving the area to scout available buildings they happen to come across. Some follow a tip from a “guy who knows a guy” about a vacancy in their desired market. Others ask a commercial real estate broker about available properties in a neighborhood they favor because they personally know it. This reliance on locally available, circumstantial information results in “starting too small.” They limit themselves in a way that statistically rules out much better location options from the very start.

The More Profitable Alternative: Companies should start by evaluating several potential markets mathematically in order to prioritize the top markets as a whole. Then they should systematically break each market into equally sized, small zones for profitability analysis. Only then, having identified the most profitable neighborhoods, should they identify particular properties for closer inspection.

Mistake 2: Falling victim to confirmation bias

Background: The human mind gets easily overwhelmed when it faces too many options, regardless of a person’s intelligence or experience. To avoid this cognitive overload, many site selection teams narrow their universe of options prematurely and with the wrong methodology. That universe is large: the United States has many markets (metropolitan statistical areas), and each has dozens of neighborhoods or zones.

The Big Costly Mistake: When companies stumble upon candidate locations circumstantially, they tend to fixate on this tiny subset as if it represented the entire universe of choices. This is what statisticians call finding a “local maximum.” The term comes from optimization work: a search limited to a very narrow domain selects the best alternative from that domain, not the global maximum that a wider search might have found.

Why do so many companies then grow more confident over time that they have chosen correctly? Because we are all susceptible to confirmation bias, a cognitive science term for the tendency to favor information that we already have or have become attached to. Humans unconsciously work to find evidence that confirms what they already secretly favor.

The More Profitable Alternative: A neighborhood-based approach that takes an entire market or set of markets into account mathematically cuts down on or eliminates confirmation bias. The profitability score for each neighborhood speaks for itself, and ranking the zones by profitability removes the chance for bias to influence which specific locations the team evaluates.

Mistake 3: Failing to distinguish between neighborhood factors and site-specific factors

Background: Two main sets of factors contribute to a candidate location’s profit potential. Neighborhood factors relate to who lives and works in the neighborhood: your potential customers, and how likely they are to buy from your new location. Site-specific factors are something else altogether: how easy it is to enter and leave the parking lot (ingress and egress), the attractiveness of the building and grounds, the type and count of other businesses in the same center, and how visible the site is from nearby major roads.

The Big Costly Mistake: Evaluating neighborhood and site-specific factors at the same time, or out of order, conflates the role that each should play in the decision. A beautiful or functional facility cannot make up for a neighborhood that fits your brand poorly. Conversely, a great neighborhood teeming with potential customers will not necessarily produce strong sales if the experience of getting into, and spending time in, the location is unfavorable.

The More Profitable Alternative: Break these two sets of factors out, treat them independently and approach the analysis in the right order. First, choose the neighborhood for profitability through mathematical evaluation and prioritization. Then, evaluate the site-specific factors for the resulting candidate locations. In that order. Every time.

Mistake 4: Failing to evaluate the trade area at the correct radius levels

Background: Choosing the right radius level, or set of radius levels, is crucial when evaluating the neighborhood factors around a set of candidate locations.

The Big Costly Mistake: Companies that use self-service SaaS tools are often unclear as to how far out to go with their radius or drive-time settings when pulling area demographic data and population counts. This can lead to statistics that take into account too much, or too little, of the surrounding geography.

The More Profitable Alternative: Ambient Array’s neighborhood factor analysis incorporates data about our clients’ historical customer base, calculating the average distance traveled for each customer. We use statistical techniques to determine the relevant radius. Then we go a step further: we pull data at multiple radius levels (for example, at 3 miles, 10 miles and 25 miles), and layer our research with statistical data about each radius level before combining them into a single revenue prediction score.

Mistake 5: Ignoring historical success drivers

Background: Looking at the data is important, but how does a team know which data matters when evaluating specific locations? The correct answer is not always intuitive.

The Big Costly Mistake: Most site selection professionals who use self-service tools make assumptions about which data dimensions represent success factors for a site. For example, many assume that having more high-income households in the area is always more desirable. But for some businesses, a healthy mix of households across multiple income levels may be a better predictor of success. The assumptions they make about the correct data to pull and analyze are often incorrect.

The More Profitable Alternative: At Ambient Array, we pre-test for statistical correlations between individual neighborhood variables and each location’s historical profitability. What better place to look for historical success factors than within your own data?

Mistake 6: Not taking enough neighborhood factors into account

Background: There are dozens or even hundreds of potential factors that can be evaluated as contributors or non-contributors to a location’s success.

The Big Costly Mistake: Many site selection teams look only at simple demographics like age, income and population counts. By focusing on such a small number of variables, they will likely overlook the factors that actually matter to their own brand’s success. They also tend to look only at absolute values (household or company counts) instead of percentage concentrations of desirable household or business types.

The More Profitable Alternative: Ambient Array starts the model-building process by analyzing dozens of potentially relevant variables, at multiple radius levels, for each of your existing locations. We take into account not only simple demographics like age and income, but also dozens of psychographic and commercial factors. We use both absolute values and percentages. And we test them for relevance, so not all of the factors make it into the final model.

Mistake 7: Combining relevant site factors in a simplistic manner

Background: Once a team has identified the significant variables, it needs a way to combine them into a single predictive score, either a relative index score or a revenue prediction score.

The Big Costly Mistake: Site selection teams often use simple additive or averaging methods to combine these variables. They also often give each dimension the same weight, even though the various factors contribute to success to varying degrees. This can lead to a model that is far less predictive than it could be.

The More Profitable Alternative: Ambient Array uses neural network technology, specifically supervised machine learning methods, to combine the final variables into a predictive model. We feed the relevant variable data, along with historical revenue performance, into the model. We run it for thousands of cycles until it has trained itself on the combination of variables and weights that best predicts revenue. Once the model is built, we run the same variables (minus historical revenue, which by definition is not yet known) through it to produce a prediction score for candidate locations.

Mistake 8: Placing the burden of analysis on non-data scientists

Background: Accessing the right data, pulling raw location-specific data, normalizing it, isolating the significant variables and combining them into a predictive model is a multi-step process. It is best left to professional data and statistics specialists who also have deep subject matter expertise in real estate and marketing.

The Big Costly Mistake: Self-service tools give site selection teams a false sense of security. When users pull the data for the neighborhood around an address themselves and read the results on the screen, they can be misled into concluding that a proper evaluation has been made on that location.

The More Profitable Alternative: Model-based approaches involve a process. They are not plug-and-play. By hiring a model-building data science team, the user gives up the immediate satisfaction of plugging an address into a software tool. What they gain is a measurably more accurate, predictive model that they can feel confident passing to the CEO, CFO or board for final approval of the neighborhood. There is also little work for the team at the neighborhood analysis level, so they can focus their time on site-specific factors for a small handful of candidate locations.

Three more pitfalls: accessibility, bias and cannibalization

Ignoring site accessibility

Once the data has narrowed your options, drive the area, then walk it. Be wary of becoming enamored with a location without taking a sober look at on-site factors. The one many teams overlook is accessibility, also known as ingress and egress.

We had a client in the early 2010s who had chosen a location for their large venue entertainment concept adjacent to a sizable shopping mall. The foot traffic from the mall seemed promising, and the location scored high on visibility, with signage visible from a nearby, well-traveled highway. The catch only became clear after they signed the lease: people could see the location from the highway, but getting there was a nightmare. The venue was tucked behind the mall, and reaching it required intimate knowledge of the shopping center’s labyrinthine parking lot. The location’s performance suffered significantly.

Do not rely on maps or satellite images. Observe car and foot traffic around the location and on the connector roads between major highways and the entrance, and ask:

  1. How easy is it to spot the entrance from main roads?
  2. Are there clear, logical routes to reach the location?
  3. Is there potential for traffic congestion during peak hours?
  4. For pedestrians, are there safe, convenient paths to your door?

Even an attractive location with the right demographic mix can fail if customers find it frustrating or difficult to reach.

Selection bias

Confirmation bias (Mistake 2) is not the only bias that distorts location decisions. Three others do as well:

  • Availability bias. Recent experiences get undue weight. After one highly successful urban development, decision-makers might overestimate similar urban locations.
  • Anchoring bias. The first site viewed becomes the reference point. If it has excellent foot traffic but high rent, decision-makers might overemphasize foot traffic in every later evaluation.
  • Overconfidence bias. Years of successful decisions can lead experienced professionals to trust their gut feelings over hard data.

Mathematical models counter these biases. Every location is evaluated on the same criteria, set before anyone views a property, so personal preferences and first impressions have less room to skew the analysis. Models draw on large amounts of historical and market data, not just recent experience, and they give objective feedback on prediction accuracy. They should complement human judgment, not replace it.

Cannibalization

Cannibalization occurs when a new location draws customers away from an existing one instead of attracting new customers. Picture a new coffee shop just blocks from your current successful one. If it primarily serves your existing customers, you are competing with yourself: the same sales spread across more locations, with higher operating costs.

Three ways to assess the risk:

  • Trade area analysis. Map the region from which each existing store draws its customers, then look for overlap with proposed sites. If a supermarket draws customers from a 5-mile radius and the chain is considering a new location 3 miles away, the overlap signals a high risk, unless demographic data shows rapid population growth or underserved segments within it.
  • Customer overlap evaluation. Use transaction data, loyalty program information and mobile device data to see how many customers visit multiple locations. A healthy multi-store network typically shows minimal overlap.
  • Target audience identification. Aim new locations at distinct customer segments. A fashion retailer whose mall store serves mainly weekend shoppers and younger consumers might open in the business district to capture professional shoppers.

Different formats can help as well: a restaurant chain might run a full-service location in one area and a quick-service format in another. The risk can never be completely eliminated, and some chains accept it on purpose, as with the clustering strategy described earlier.

Software, data sources and reports

GIS and mapping software. Geographic information systems (GIS) combine multiple data layers, from basic topography to demographic overlays, into a visual representation of potential locations. Solutions like ESRI’s ArcGIS and QGIS offer features for analyzing trade areas, calculating drive times and reporting on population density, median income and the competitive landscape within specific radius rings.

Data sources. The critical categories are demographic data, mobile analytics (foot traffic, customer movement and dwell times), economic indicators (employment rates, business growth trends and commercial development plans) and competition analysis (competitor locations, market share and performance metrics).

Reports. Three report types support the decision:

  • Market potential reports combine demographic data with spending patterns to estimate market size and growth potential.
  • Site comparison matrices set multiple locations side by side, weighing factors like accessibility, visibility and market potential.
  • Risk assessment reports analyze the potential risks at each location, from environmental concerns to market saturation.

As Mistake 8 explains, having access to software and data is not the same as having a proper evaluation.

How site selection services are priced

Ambient Array structures its site selection service in two components:

  • Model build. A one-time investment that establishes your customized site selection model. During this three-week process, we work with your team to understand your requirements and business metrics and develop the model behind all future site evaluations.
  • Site evaluation. Evaluations run in batches. Each batch can include 10 to 30 locations and is typically completed within 5 to 7 business days.

Separating the model build from ongoing evaluations means you pay only for the analysis you need, when you need it. The model can be applied repeatedly as your business grows, and batches let you assess new locations on your own timeline and budget.

Getting started

The short version: start with markets and neighborhoods, not properties. Let your own historical data decide which factors matter. Evaluate on-site factors, especially accessibility, after the neighborhood analysis. And check new sites against the locations you already operate.

To have a custom model built for your organization, see Ambient Array’s real estate site selection service or contact Ambient Array.

Start the Conversation

Ready to obtain strategic clarity for your marketing machine—or to build a new one from scratch? Start the conversation with (human-powered, data-infused) Ambient Array today.