Network modeling and site scoring are often treated as competing frameworks in logistics real estate decisions, but practitioners who default to one at the expense of the other consistently arrive at flawed shortlists. Each method answers a different question, and the real analytical work begins when both are run in sequence, with the outputs of network modeling informing the criteria weights that drive site scoring. This article explains how each method works, where each breaks down in isolation, and how to design a combined evaluation process that produces defensible, executable real estate decisions.
What Network Modeling Actually Measures
Network modeling is a supply chain optimization discipline, not a real estate tool. Its core output is a recommended configuration of nodes — distribution centers, cross-dock facilities, fulfillment hubs — positioned to serve a defined demand pattern at minimum total cost. The demand pattern includes customer locations, order frequencies, and service-level commitments such as next-day or two-day delivery zones.
The cost inputs in a network model typically include transportation rates by lane and mode, facility operating costs at assumed square footages, labor costs by geography, and inventory carrying costs tied to network structure. The model optimizes across these inputs to identify how many facilities are needed, approximately where they should be located, and what size each should be.
What the model does not produce is a specific address or a ranked list of available buildings. It produces a geographic zone — sometimes as broad as a metropolitan statistical area, sometimes as narrow as a county or zip code cluster — where a facility should theoretically exist. That output is a constraint, not a destination.
The critical insight for real estate practitioners is that the zone is probabilistic, not deterministic. Shift a transportation cost assumption by ten percent, and the optimal zone may shift by fifty miles. This sensitivity is not a deficiency in the model; it is an honest representation of how supply chains actually behave under changing market conditions. Building real estate decisions on a single network model run without sensitivity analysis is a known failure mode.
What Site Scoring Actually Measures
Site scoring is a structured comparative evaluation of specific properties against a defined set of criteria. The criteria are weighted, each candidate property receives a score on each criterion, and the weighted scores are summed to produce a composite ranking. The ranking orders properties for further diligence and negotiation.
The criteria in a logistics site scoring exercise typically span several categories. Labor supply and wage rates, building specification, zoning and permitting history, truck court dimensions, transportation access, utility capacity, and lease economics are common categories. Each category may contain multiple sub-criteria, and the weighting reflects how the occupier's operation ranks those factors relative to one another.
Site scoring answers a question that network modeling cannot: given a pool of available properties within or near the optimal zone, which specific building offers the best combination of operational fit and economic value? The answer depends on current market inventory, landlord flexibility, and the occupier's internal priorities — none of which appear in a network model.
The limitation of site scoring in isolation is that it has no mechanism to test whether the geographic cluster of candidate properties actually serves the occupier's demand efficiently. A building may score perfectly on labor, specification, and lease economics, but if it sits thirty miles outside the zone identified by the network model, it may add days to delivery times or millions of dollars in annual transportation cost. Scoring without modeling produces locally optimal choices that may be globally suboptimal.
The Sequencing Logic: Model First, Then Score
The combined methodology runs in a clear sequence. Network modeling establishes the geographic constraint — the zone or set of zones where facilities should be located. Site scoring then evaluates specific properties within those zones. The two methods are not parallel tracks; they are sequential stages with a defined handoff.
The handoff point is the moment when the network model output is translated into a geographic search boundary for the real estate team. That translation requires judgment. If the model identifies a broad zone, the real estate team needs to understand which sub-markets within that zone offer adequate building inventory, labor access, and infrastructure before they begin pulling properties for scoring. A zone that looks attractive on a transportation cost map may contain no suitable industrial product at any realistic construction timeline.
Practitioners who run both methods well treat the network model output as a living constraint, not a fixed answer. When the real estate search reveals that the optimal zone has insufficient inventory or prohibitive lease economics, the finding feeds back into the network model as a sensitivity case. The modeler reruns the optimization with adjusted assumptions — higher occupancy cost in the preferred zone, lower cost in an adjacent market — and the team evaluates whether the network cost increase from shifting the zone is offset by real estate savings.
This iterative loop is where most occupier teams lose efficiency. Without a structured process for passing findings between the network modeling team and the real estate team, each group tends to treat the other's output as a black box. The network modelers hand over a zone and consider their work complete. The real estate team selects a building in that zone and considers theirs complete. Neither team tests whether the final decision is consistent with the original network cost assumptions.
Defining the Zone from Network Output
Translating a network model output into a usable real estate search boundary requires several steps that are often skipped or underspecified. The first step is identifying the sensitivity range of the optimal zone. A well-run network model includes a sensitivity table showing the total cost impact of locating the facility at various points across the feasible region. That table defines how much geographic flexibility the occupier has before network costs rise materially.
If the sensitivity table shows that locating within a thirty-mile radius of the optimal centroid adds less than two percent to total annual network cost, the real estate team has meaningful geographic flexibility. If it shows that moving twenty miles in any direction adds eight percent to annual network cost, the zone is tight and the real estate search must concentrate heavily in a small sub-market.
The second step is overlaying real estate market data onto the sensitivity range. This means pulling available industrial inventory by size, specification, and lease term within the feasible geographic band, and noting where current vacancy concentrates. Markets with low vacancy require longer lead times and may force a build-to-suit conversation; markets with higher vacancy offer more negotiating leverage on lease economics and improvement allowances.
The third step is flagging infrastructure constraints within the feasible band. Highway access, intermodal rail connections, port proximity for import-dependent supply chains, and power and water capacity for facilities that require significant utilities all narrow the practical search boundary further. A logistics real estate search that ignores infrastructure constraints during zone definition will produce a shortlist that fails technical diligence later.
Building the Scoring Matrix for Logistics Properties
A scoring matrix for logistics site selection is not a universal template. The weights assigned to each criterion must reflect the specific operation the occupier is running. A cold-chain operator weights power reliability and dock door density differently than a parcel fulfillment operation. A building used for regional replenishment of retail stores has different labor demand than a last-mile facility serving residential customers.
The process for setting weights begins with a structured session involving the occupier's supply chain, real estate, finance, and operations teams. Each team brings different priorities. Supply chain focuses on transportation access and network fit. Operations focuses on building specification and labor availability. Finance focuses on occupancy cost and lease term flexibility. Real estate focuses on market conditions and landlord capability. The scoring matrix must reflect a negotiated set of priorities, not a single function's preferences.
Criterion weights are typically expressed as percentages summing to one hundred. A logistics site scoring matrix might allocate thirty percent to transportation and logistics access, twenty-five percent to labor supply and wage rates, twenty percent to building specification and infrastructure, fifteen percent to lease economics including effective rent and improvement allowances, and ten percent to regulatory and permitting factors. Those allocations are illustrative; every occupier's situation will produce different weights.
Scoring each criterion requires a consistent scale. A five-point or ten-point scale works well for most exercises. What matters is that the scoring rules are written down before properties are evaluated, so that the same standard is applied across all candidates. If "highway access" scores five points for direct interchange access within one mile, that definition must be fixed before the team begins pulling property data, not adjusted retroactively to favor a preferred building.
Handling Labor in the Combined Framework
Labor is consistently one of the most consequential and least modeled inputs in logistics site selection. Network models typically use aggregate wage rates by metropolitan area, which masks significant variation within a market. A facility located on the east side of a metro area may draw from a substantially different labor pool than one on the west side, even if both fall within the optimal zone identified by the network model.
The scoring matrix must disaggregate labor into components that reflect the occupier's actual hiring profile. Total addressable workforce within a defined commute time, wage rates for the specific job categories the facility will hire, turnover rates in comparable facilities in the area, and the competitive hiring environment — including how many similar facilities already operate in that labor shed — are all distinct criteria that aggregate wage data does not capture.
Labor analysis also has a timing dimension that neither network modeling nor static scoring handles well. A labor market that appears favorable at the time of site selection may tighten significantly by the time a build-to-suit facility opens or a lease commences. Practitioners who include a forward-looking labor assessment — examining planned facility openings in the same labor shed, regional workforce development programs, and population migration trends — make more durable location decisions.
The handoff between network modeling and site scoring is particularly important for labor because the network model's geographic zone may bisect a metro area's labor market in ways that are not obvious from a transportation cost map. A zone that appears geographically central may fall in an area with limited transit access, reducing the effective labor pool for shift workers who do not own vehicles. That constraint only surfaces when the real estate team maps available properties against transit routes — a step that belongs in the site scoring process, not the network model.
Integrating Lease Economics into the Scoring Framework
Lease economics deserve their own scoring dimension because the total occupancy cost of a logistics facility has a direct relationship to network cost. A building that is twenty miles outside the optimal zone but carries a substantially lower net rent may be cost-neutral or even cost-favorable once transportation differentials and occupancy savings are compared on a net present value basis. That comparison requires the real estate team to run a lease NPV calculation for each candidate and feed the results back into the network model as an adjusted occupancy cost assumption.
Effective rent — the average rent paid over the lease term after accounting for free rent periods, tenant improvement allowances, and rent escalations — is the standard metric for comparing lease economics across candidates with different structures. A building offering eighteen months of free rent on a ten-year lease may have a lower effective rent than a building with a lower face rate but no free rent and aggressive annual escalations. Scoring on face rate alone systematically misleads the evaluation.
Improvement allowances deserve separate treatment in a logistics context because the gap between base building condition and the occupier's operational requirements can be substantial. Clear height, dock door count and configuration, trailer parking, electrical capacity for automation, and floor flatness tolerances all affect how much improvement work is required. A building with a generous allowance but significant improvement requirements may deliver less net benefit than a building with a smaller allowance and a specification that is already close to fit-for-purpose.
Running Sensitivity Analysis Across Both Methods
Sensitivity analysis is the mechanism that connects the two methods into a coherent decision framework. The network model produces sensitivity outputs showing how total cost changes as facility location shifts across the feasible region. The site scoring process produces sensitivity outputs showing how the composite ranking changes when criterion weights are adjusted. Combining both reveals which decisions are robust — meaning they hold across a wide range of assumptions — and which are fragile.
A robust decision is one where the leading candidate from site scoring also falls comfortably within the low-cost zone from the network model, and where the site scoring ranking is stable even when criterion weights are shifted significantly. That combination provides strong confidence that the selection will perform well under real operating conditions.
A fragile decision is one where the leading candidate from site scoring is near the edge of the feasible network zone, and where small changes in transportation cost assumptions or criterion weights would change either the zone boundary or the scoring rank. Fragile decisions are not automatically wrong, but they require additional scrutiny — typically a deeper dive into the specific transportation lanes that make the zone boundary sensitive, and a stress test of the scoring criteria that are driving the candidate's composite rank.
The practical question practitioners face is how many sensitivity cases to run before a decision is sufficiently tested. A minimum reasonable standard is to test the network model with transportation costs at current rates plus a fifteen percent increase, and to test the scoring matrix with the top two criterion categories each weighted ten percentage points higher than the base case. If the leading candidate holds its position across all four scenarios, the decision is reasonably robust. If the ranking changes materially in any scenario, the team needs to understand why before proceeding.
Documenting the Combined Analysis for Stakeholders
Real estate decisions of this scale — a logistics facility representing a multi-year lease commitment and significant capital for fit-out — require documentation that allows senior decision-makers to understand and challenge the analysis. A well-structured decision document explains the network model methodology and key assumptions, the geographic zone derived from the model, the criteria and weights used in site scoring, the scores assigned to each candidate and the reasoning behind them, the lease economics comparison, and the sensitivity analysis results.
The documentation requirement is not bureaucratic overhead; it is a risk management discipline. When a decision is challenged six months after signing — because a competitor opened a facility that disrupts the labor market, or because transportation costs shifted in a way that changed the network economics — the team needs to show that the analysis was sound given the information available at the time. Undocumented decisions cannot be defended on those terms.
Where the Combined Approach Fails and How to Prevent It
Even practitioners who understand both methods and run them in sequence encounter failure modes that undermine the final decision. The most common is the stale network model. Network models are expensive to build and time-consuming to run, so organizations often reuse a model that was built for a previous decision, updating only the most obvious inputs. A model built on two-year-old transportation rate data, pre-pandemic demand patterns, or a service-level commitment that has since changed will produce a zone that does not reflect current operational reality.
The second common failure is criterion drift in the scoring matrix. This happens when the team adjusts scores or weights during the evaluation process to steer toward a preferred outcome — usually a building that someone in the organization has already decided to pursue for reasons unrelated to the analytical framework. Criterion drift is difficult to detect after the fact, which is why the weights and scoring rules must be fixed and documented before any properties are evaluated.
The third failure mode is insufficient property inventory within the optimal zone. When the network model identifies a zone where available industrial inventory is thin — common in constrained coastal markets — the real estate team may be forced to extend the search boundary without formally retesting the network economics at the extended locations. The result is a building selected outside the feasible zone, with no analysis of the network cost implications.
Preventing these failures requires process discipline: a dated network model with a formal refresh trigger, a scoring matrix that is locked before evaluation begins, and a feedback protocol that routes real estate market findings back into the network model whenever the search boundary is extended. None of these controls require sophisticated technology, but all require deliberate process design.
Applying the Framework to Build-to-Suit vs. Existing Building Decisions
The combined methodology applies differently depending on whether the occupier is considering a build-to-suit facility or an existing building. For an existing building, the site scoring process evaluates current building condition and specification against occupier requirements, with improvement costs factored into the lease economics comparison. The network model zone either contains suitable existing product or it does not, and the team must decide how much network cost penalty is acceptable to extend the search to a market with better inventory.
For a build-to-suit, the site scoring process shifts its emphasis. Building specification becomes a design variable rather than a fixed attribute, so the scoring matrix de-emphasizes current specification and places more weight on land characteristics, zoning and entitlement risk, development timeline, and developer capability. The network model zone still defines the geographic constraint, but the real estate search now focuses on developable land parcels rather than existing buildings.
The key distinction for the build-to-suit path is timeline. A build-to-suit typically requires twelve to thirty-six months from site control to occupancy, depending on jurisdiction, building size, and specification complexity. The network model assumptions embedded in the original zone analysis may no longer be valid by the time the facility opens. Teams pursuing build-to-suit solutions should rerun the network model at the point of design finalization to confirm that the assumed operating costs and service levels still produce the same optimal location.
How the Methods Apply to Multi-Node Network Decisions
The analysis becomes substantially more complex when the decision involves configuring multiple facilities simultaneously — expanding a network from two distribution centers to five, for example, or reconfiguring an existing network in response to a channel shift from wholesale to direct-to-consumer fulfillment. In these cases, network modeling is not optional; the interdependence between node locations makes it impossible to evaluate any single site in isolation.
Multi-node network decisions require the network model to solve for the full configuration simultaneously, not sequentially. Adding the fifth node to an existing four-node network changes the demand allocated to each of the existing four nodes, which changes the transportation lanes and costs for the entire network. A sequential approach — picking each node in order — will not find the global optimum because it cannot account for these interdependencies.
The site scoring process for multi-node decisions must also account for the interaction between nodes. A building that scores well for a specific node in isolation may score less well when the team considers how the full network configuration divides demand across all nodes. This is the core reason why Network Modeling vs Site Scoring: How They Work Together in Logistics Decisions is a methodology question, not a technology question — the analytical architecture must be designed to reflect how supply chains actually function, not how real estate decisions have traditionally been made.
About Advantai
Advantai is a commercial real estate intelligence and operations platform operated by ADVANTAGE AI LLC, a Delaware limited liability company. It connects client relationships, property research, documents and financial decisions in one workspace for commercial real estate teams — advisers and brokerage teams, occupier and facility teams, and portfolio teams. The platform covers CRM and origination, requirements and site selection, Property X-Ray (an interactive 3D building workspace), financial modeling and comparison, document intelligence, transactions and diligence, client collaboration, and portfolio strategy with critical dates. The optional Super Agent upgrade adds specialist, source-backed research and automated scenario analysis.
Get Started with Advantai
Ready to see your next move clearly? Go to advantaico.com, click Request a demo and tell us about your next project. Prefer to start with a single project? Visit advantaico.com/getting-started to plan your first one.