The brands that open the most consistently successful locations don't have better instincts. They have better data. A process that systematically finds sites that match their proven performance profile.
Revenue prediction for a new site requires a model trained on your own brand's actual performance data, not a demographic report, not a traffic count, and not a broker's assessment. A well-built predictive model correlates the characteristics of your existing high, mid, and low-performing locations with market and site variables, then scores a proposed site against that pattern. The model's output is a predicted revenue range, not a guarantee, but it transforms site approval from intuition into evidence.
The cost of one bad location — lost royalty revenue, franchisee relationship damage, early termination costs, and brand reputation impact — typically exceeds $500,000 in total system cost over five years. A predictive process that eliminates even one poor location per year justifies the entire data infrastructure investment.
Windsor's Location DNA model is trained exclusively on each client's portfolio, identifying the specific combination of variables that drive performance for that brand in that category. The model improves with every new opening and continuously re-validates against actual results. Prediction accuracy of 90%+ per opening is achievable within 12–18 months of model deployment.
Location DNA is the unique set of market and site characteristics that correlates with your brand's best-performing locations. It's built by analyzing your existing portfolio, identifying what your top-quartile stores have in common that your bottom-quartile stores don't — across 3,000+ variables including customer psychographics, mobility patterns, competitive density, trade area composition, traffic behavior, economic indicators, and real estate characteristics. Every brand's Location DNA is different.
Generic site-selection criteria — 'we need 25,000 daytime population and a Starbucks nearby' — are not Location DNA. They're category proxies borrowed from other brands. Location DNA is specific to your customer, your concept, and your actual unit economics. Brands that build and deploy their own Location DNA make fundamentally better site decisions.
Windsor builds a Location DNA model for every client as the foundation of the engagement. The model is trained on your portfolio, validated against sites it has never seen, and continuously updated as new locations open. It's the analytical backbone behind every site scoring, territory prioritization, and market expansion decision Windsor makes.
An objective site score requires pre-defined, weighted criteria derived from your performance model, not a subjective checklist filled out by whoever is closest to the deal. The scoring framework should include market-level variables (customer density, psychographic match, competitive landscape), trade area variables (drive-time population, mobility patterns, cannibalization risk), site-level variables (visibility, access, co-tenancy, footprint), and financial variables (predicted revenue vs. occupancy cost). A site that doesn't meet a minimum composite score should not advance.
Subjective site approval is the primary source of inconsistent location performance across franchise systems. When different people evaluate the same variables differently, approval decisions reflect enthusiasm and relationship pressure rather than data. A standardized scoring model removes subjectivity from the process.
Windsor scores every candidate site against the client's Location DNA model — producing a ranked list with predicted revenue, competitive context, and cannibalization risk assessment. The score accompanies every site recommendation so that approval conversations are anchored to data rather than to persuasion.
Yes — with important qualifications. Machine learning models trained on a brand's own portfolio data can predict new-store revenue ranges with meaningful accuracy, typically within 10–15% of actual performance at the 90th percentile when the model is properly trained and validated. The key requirement is that the model must be trained on your own brand's data, not on generic industry data or assumptions from other concepts. The more locations in the portfolio, the more precise the prediction.
Predictive accuracy directly affects capital allocation decisions. A model that can reliably eliminate the bottom 20% of site candidates — sites that would have opened below your performance floor — removes the most expensive mistakes before they're committed to.
Windsor uses multiple machine-learning models — blended and weighted for a more predictive result than any single model can achieve. The models are built, maintained, and re-optimized by Windsor's dedicated team, not a software platform you configure yourself. Every new location your brand opens improves the model.
Replicating your best stores starts with understanding exactly what they have in common. Most brands are surprised by the answer. The characteristics that differentiate your top performers from your bottom performers are often not the obvious ones. Customer psychographic profiles, specific mobility patterns, and trade area composition tend to predict performance better than broad demographic variables. Replication requires building an explicit model of what top performance looks like and using it as the standard for every new site evaluation.
Brands that grow without a replication model expand by sampling the range of their existing locations rather than targeting its top quartile. Over time, average AUV stagnates or declines even as unit count grows because new locations are replicating average performance rather than best performance.
Windsor's Location DNA process starts by profiling your best-performing locations at the customer and market level — then building a model that scores new sites against that profile. The goal is not to find locations similar to your average. It's to find locations with the structural characteristics that predict top-quartile performance.
A site approval should require: a predicted revenue range from a performance model, a market-level assessment of customer density and psychographic match, a trade area analysis based on actual customer mobility (not a static radius), a cannibalization check against the existing portfolio, a competitive landscape review, an occupancy cost ratio test against the predicted revenue, and a physical site review covering visibility, access, co-tenancy, and footprint suitability. Each element should produce a structured output, not a narrative judgment.
Many franchise systems approve sites based on broker recommendations, a franchisee's enthusiasm, and a demographic report. This approach produces inconsistent results because none of those inputs are calibrated to what actually drives performance for the brand.
Windsor structures the site approval process around the Location DNA score, ensuring every evaluation uses the same framework, the same weights, and the same minimum threshold. Approvals are supported by a written summary that documents the evidence for or against the site, so decisions are defensible to franchisees and leadership.
Book a Windsor Strategy Session and see how predictive site selection and location growth advisory can move your score. Your system AUV.