Route Optimization for Enterprise 3PLs: Why Multi-Client Routing is a Cost Allocation Problem

Route optimization for a 3PL is a different problem from route optimization for a shipper. A shipper optimizes one network against one set of service commitments. A logistics service provider optimizes one network against many clients, each with its own windows, vehicle requirements, proof-of-delivery rules and contractual penalties, while co-routing their stops together because density is where the margin lives. That combination makes 3PL routing a joint optimization and a cost allocation problem at the same time, and most platforms solve only the first half.

Why 3PL Routing Differs from Shipper Routing

Density is not a nice-to-have in this business, it is the business. McKinsey’s work on last-mile delivery finds the last mile generates 60 to 70 percent of overall parcel delivery costs, and identifies drop density as the main driver of unit economics, noting that raising parcels delivered per stop from one to five materially changes the economics. A 3PL’s structural advantage over a shipper running its own fleet is precisely that it can assemble density a single shipper cannot.

Cost pressure makes capturing that advantage more urgent than it was. ATRI’s operational costs of trucking analysis put the industry average at $2.336 per mile in 2025, the highest in the report’s history, with increases across every major line item. In a business running on single-digit margins, the difference between routing clients together and routing them apart is often the difference between a profitable contract and a loss-making one.

The complication is that clients do not buy density, they buy service. Each one holds a contract with its own delivery windows, vehicle and handling requirements, proof-of-delivery standards, reporting formats and penalty structures. Co-routing them means one vehicle simultaneously satisfying several contracts, then producing separate measurement and separate invoices from a shared operation. The routing engine decides how the cost falls, whether or not anyone configured it to.

Doing this well depends on data from several systems at once, which is where it usually breaks. Gartner’s survey conducted in October and November 2025 found that more than half of chief supply chain officers, 56% of those surveyed, cite integrating AI with legacy systems as a major challenge. Per-client attribution requires order data, execution data and cost data to agree, and MIT Sloan Management Review puts the cost of bad data at 15% to 25% of revenue for most companies. A 3PL with imperfect attribution is not merely reporting inaccurately, it is pricing inaccurately.

Also Read: Fleet Utilization for 3PLs: Multi-Client, Multi-Fleet Operations

How to Route and Allocate Across Clients

1. Co-route for density, then attribute deliberately

The sequence matters. Optimize the network across all clients to capture the density dividend, then allocate the resulting cost using a rule you can defend to a client who asks. Reversing the order, by allocating first and optimizing within client boundaries, forfeits the dividend to protect the accounting, which is the wrong trade for a business whose advantage is density. This is worth checking rather than assuming, because many 3PLs route by client without having decided to. It is the default that emerges when each account manager owns a plan and no one owns the network.

2. Measure the density dividend before arguing about who owns it

Work a concrete case. Client A has 40 stops in a tight urban cluster and would cost $800 to serve alone. Client B has 10 stops scattered across the same territory and would cost $400 alone. Co-routed, the combined route costs $950.

The standalone total is $1,200 against a joint cost of $950, so the density dividend is $250, or 20.8% of standalone cost. That number is the prize. Every allocation argument that follows is a negotiation about who receives it. Calculate it on real routes before designing a rate card, since the size of the dividend determines how much room there is to price competitively without giving away margin.

3. Understand that the allocation method is a pricing decision

Four methods are in common use and they produce materially different bills from an identical operation.

Client B’s bill ranges from $150 to $317 across those methods, a spread of 111% on the same route with the same vehicle on the same day. Nothing operational distinguishes the four cases. The difference is entirely a choice about accounting, and in most 3PLs it is a choice nobody made explicitly.

4. Know why the most common method is the least defensible

Stop-count proportional allocation is the industry default because it is easy to explain and easy to compute. It is also the method that misprices most severely.

In the worked example, B contributes 20% of the stops and receives 84% of the density dividend. The reason is visible in the standalone numbers: B’s cost per stop alone is $40 against A’s $20, because B’s stops are scattered and A’s are clustered. B is the expensive client. Charging by stop count prices the expensive client as though it were the cheap one, and the high-density client subsidizes it.

That is how a 3PL ends up with a large, apparently profitable client that is actually being carried by a smaller one, and no report shows it because the allocation rule generating the report is the thing causing the distortion. The tell is a client whose reported cost per stop is close to the network average despite stops that are visibly scattered on a map.

5. Use the method that has a proof behind it

Cost sharing among parties who create joint savings is a solved problem in cooperative game theory. The Shapley value, introduced in Lloyd Shapley’s 1953 work on n-person games, allocates cost by averaging each participant’s marginal contribution across every possible order in which participants could join the coalition. It is the unique allocation satisfying anonymity, full cost recovery and strong monotonicity, which in plain terms means it depends only on the cost impact each client actually has, it allocates the whole cost with nothing left over, and a client whose contribution rises never sees its allocation fall.

Applied to the example, Shapley gives A $675 and B $275, splitting the $250 dividend evenly at $125 each. Extended to three clients, the divergence from stop-count allocation widens further.

Three clients co-routed cost $1,150 against $1,700 standalone, a 32% dividend. Both methods recover the full $1,150, but stop-count overcharges A by 13% and undercharges B by 30%. At contract scale those percentages are the difference between a renewal and a re-tender. Computing Shapley across many clients becomes expensive quickly, since the number of coalitions grows exponentially, so production implementations typically use sampled approximations. The approximation is still closer to defensible than stop count.

6. Never quote marginal cost to more than one client

Marginal-cost pricing is intuitive and mathematically guaranteed to lose money in a shared network. Each client’s marginal cost is what it adds to a route that already exists, so B’s marginal cost is $150 and A’s is $550. Those sum to $700 against an actual cost of $950.

The $250 shortfall is exactly the density dividend. Quoting incremental cost to win each deal individually means the dividend is given away twice, once to each client, and the 3PL absorbs it. This is the most common pricing error in competitive 3PL bidding and it is invisible until the contracts are all live.

7. Account for SLA transfer between co-routed clients

Cost is not the only thing shared. If Client A requires a 9am to 12pm window and Client B accepts same-day by 6pm, the shared route is bound by the three-hour window. B’s stops are now served under a constraint B did not buy and is not paying for, and A’s tight window is subsidized by B’s flexibility.

This runs in both directions and it is worth pricing. A client with genuinely flexible windows is contributing something real to network efficiency, and a rate card that ignores window width prices flexibility at zero. Pricing it explicitly also gives the commercial team something to trade, since a client resisting a rate increase will sometimes accept a wider window instead.

Also Read: Smart 3PL Delivery Orchestration: Complete Guide

Shipper Routing and 3PL Routing Compared

The final row is the distinction that matters. A 3PL can run an excellent optimizer, capture a full density dividend, and still lose its most profitable client because the invoice cannot be explained. Routing quality and attribution quality are separate capabilities, and a platform that delivers only the first leaves the commercial half of the problem on the operator’s spreadsheet.

Five Things to Evaluate in a 3PL Routing Platform

1. Multi-client co-routing with per-client measurement. Confirm the system can build one route across several clients while measuring each client’s SLA against that client’s own definition of on-time. Ask to see one route producing several separate client reports.

2. Configurable cost allocation. Ask which allocation methods are supported and whether the rule can differ by contract. A platform that outputs only total route cost has handed you the harder half of the job.

3. Per-client constraint modeling. Each client brings vehicle types, handling requirements, POD standards and access rules. Establish whether these are modeled as hard constraints inside the optimization or applied as filters afterward, since filtering after the fact forfeits density.

4. Attribution data at execution level. Allocation is only as good as the underlying record. Confirm the platform captures actual distance, service time and stop sequence per client on shared routes rather than estimating them from the plan.

5. Impact modeling for new client onboarding. Adding a client re-optimizes the network. Ask whether the system can simulate the effect on existing clients’ routes and service before the contract is signed, because discovering it afterward is a renewal conversation. The same simulation answers the pricing question, since it produces the density dividend the new client would create and therefore the floor beneath which the bid stops being worth winning.

Also Read: Real-Time Tracking and Visibility for 3PLs in 2026

What This Looks Like in Practice

Utilization as the shared-network payoff. Locus reports fleet utilization improving by 30% through better load consolidation and dynamic rebalancing across delivery zones. Consolidation across clients is the mechanism, and it is only available to an operator willing to co-route. The 30% figure is the density dividend expressed as an asset metric rather than a cost one, which is often the easier version to take to a client, since a utilization gain reads as competence while a cost saving invites a request for the saving.

Execution rate at multi-client scale. A Fortune 50 operation running more than 4,500 drivers moved execution rate from 75% to 92% and surfaced more than $14M in annualized operational opportunity. At that scale the routing decision and the attribution decision are made hundreds of thousands of times a day, which is why neither can be handled manually.

Planning cycle time and the re-quote problem. Locus customers connecting warehouse readiness signals to automated dispatch have reduced planning cycle time by 66%. For a 3PL this matters commercially as well as operationally, because a faster planning cycle is what allows a network impact assessment to be run during a bid rather than after it.

Four Mistakes in 3PL Route Optimization

Routing each client in its own silo to keep the accounting simple. This protects the invoice by forfeiting the density dividend, which is the only structural advantage a 3PL holds over a shipper’s private fleet.

Letting the default allocation rule set your client profitability. Stop-count allocation is a choice with a 111% swing attached. If nobody picked it deliberately, the margin report is describing an accounting artifact rather than the business.

Quoting marginal cost in competitive bids. It wins deals and under-recovers by exactly the dividend. The shortfall only becomes visible once every contract in the shared network is live.

Pricing window width at zero. Clients with flexible windows subsidize clients with tight ones on shared routes. A rate card that does not distinguish them is transferring value between customers without telling either, and the client being subsidized is usually the one with the sharper procurement team.

Also Read: 3PL TMS Integration Depth: API-First vs Legacy Platforms

How Locus Supports Multi-Client 3PL Routing

Locus, the world’s first Decision-Intelligent, Agentic TMS, models 250+ real-world constraints simultaneously, which is the property that makes co-routing across clients feasible rather than theoretical. Per-client vehicle requirements, service windows, handling rules and access restrictions enter the same optimization, so a route serving several contracts is built to satisfy all of them rather than assembled and then checked. Where clients are co-routed for density, stops can be planned together while remaining separately measured, billed and reported against each client’s own service definition.

Three dispatch levers move utilization in a multi-client fleet: load consolidation at the point of assignment, vehicle class matching against load profile, and continuous intraday rebalancing. The first is the one that produces the density dividend, and the third is what protects it when the day moves, since a shared route that degrades after dispatch damages several client relationships at once rather than one.

On the attribution side, the requirement is execution-level data rather than planned data, and six governance mechanisms covering explainability, traceability, evaluation, autonomy levels, execution sandbox and human-in-the-loop mean each decision and its resulting execution record can be traced back to the state that produced it. That record is what any allocation method needs as an input. Locus does not prescribe a cost-sharing formula, and the choice between Shapley, standalone-proportional and stop-count remains a commercial decision, but the per-client execution data that makes the calculation possible is the part that has to exist in the platform rather than in a spreadsheet.

Locus runs at 1.5B+ deliveries across 360+ enterprise customers in 30+ countries at 99.99% uptime, orchestrating 1,000+ pre-integrated carriers across owned, contracted and gig capacity. Locus has been recognized by Gartner for seven consecutive years across multiple research categories, appears in the 2026 Gartner Hype Cycle for AI-powered logistics, features ShipFlex as a Representative Vendor in the 2026 Gartner MCPMS Market Guide, holds Leader designation in the QKS SPARK Matrix for Transportation Management Systems, and ranks #1 on G2 for Route Planning software.

In October 2025, Ingka Investments, the investment arm of Ingka Group, the world’s largest IKEA retailer, acquired Locus. Locus continues to operate independently.

To model your own density dividend and per-client attribution across a shared network, schedule a demo.

Also Read: 11 Best Routing Software for Enterprise Logistics in 2026