Orchestration Makes the Network Better and Every Scorecard Worse

Logistics automation removes manual effort from individual steps, and logistics orchestration coordinates those steps so one decision accounts for all of them. The second is where enterprises stall, and the usual explanation is that the systems are fragmented or that the culture is territorial. Both can be true and neither is the binding constraint. The binding constraint is arithmetic: a plan that is optimal for the network is worse than the local optimum for every function it spans, so each function is being asked to accept a measurable degradation on the number it is judged by, in exchange for an improvement on a number nobody owns.

That is not resistance to change. It is a correct response to the incentive structure, and no amount of integration work fixes it. Locus, the world’s first Decision-Intelligent, Agentic TMS, computes the joint plan across dispatch, routing, carrier allocation and settlement in one decision layer, which makes the trade-off visible and quantified rather than argued, though deciding who absorbs it remains a management choice.

Why the jointly optimal plan degrades every local number

Take a single day’s outbound plan spanning three functions with three different objectives. The warehouse is measured on pick efficiency, transport on cost per drop, customer service on promise attainment. Each can produce a plan that is optimal for its own objective, and each such plan is worse for the others, because the sequencing that makes picking efficient scatters geography, and the grouping that makes routing efficient scatters picks.

Index every function’s own best to 100 and the picture is stark.

The bottom row is the right answer for the company. It is also the row in which nobody wins.

The network total falls from 337, the best any single-function plan achieves, to 321. That is a 4.7% improvement for the enterprise, delivered by making three people’s numbers worse by six to eight percent each. Every one of them can accurately report that the orchestrated plan degraded their performance, and every one of them will be right.

This is why the standard diagnosis misfires. Told that silos are the problem, enterprises respond with integration projects, shared dashboards and cross-functional forums. Those are useful and they do not change the arithmetic, because the arithmetic is a property of having several objectives rather than of having several systems. A perfectly integrated stack computing a joint plan produces exactly the same set of degraded scorecards.

The stakes are not small at current cost levels. ATRI’s 2026 report puts the industry-average cost of operating a truck at $2.336 per mile in 2025, a record for the series and 3.4% above the prior year, so a 4.7% network improvement forgone because nobody would accept a local degradation is a material sum on any meaningful volume.

The problem also crosses company boundaries, where it gets harder rather than easier. Armstrong & Associates put US 3PL gross revenues at $323.4 billion in 2025 against net revenues of $138.2 billion, and when a function sits inside a provider rather than inside your organization, the degraded number belongs to a different company with its own contract. Asking a carrier to accept a worse cost per drop for your network benefit is a commercial negotiation, not a management instruction. Scale makes that harder rather than easier: the American Trucking Associations reports almost 580,000 active US motor carriers as of June 2025, of which 91.5% operate 10 or fewer trucks, so the counterparties absorbing your local degradations are frequently small operators with thin margins and no capacity to fund someone else’s network optimization.

How to make the joint plan acceptable

1. Quantify each function’s concession before asking for it

The conversation fails when the ask is qualitative. Compute what each function gives up under the joint plan against its own optimum, in its own units, and put the three numbers on one page alongside the network gain. Eight percent on pick cost is a negotiable figure. A request to think about the bigger picture is not.

In practice this means running the planner four times: once optimized for each function in turn, and once jointly. Three of those runs will never be executed, which is exactly why they get skipped, and they are the only way to establish what each function’s own best actually looks like on the same day’s orders. Without that baseline there is no way to distinguish a concession from ordinary variation, and the conversation reverts to whose intuition is louder.

2. Name the owner of the number that improves

Total cost to serve improves and almost nobody is measured on it. Find out who is, and if the answer is nobody, that is the finding. A gain with no owner has no advocate in any forum where the local degradations will be raised, which is the structural reason these initiatives lose. The practical test is to ask whose bonus moves if total cost to serve falls four percent. If three function heads can each answer no, the program has three opponents and no sponsor, and that imbalance will decide the outcome regardless of how good the plan is.

3. Choose one of three measurement fixes, deliberately

There are only three that work. Move the functions onto a joint metric so their scorecards track total cost to serve rather than local efficiency. Hold them harmless by reporting both their local number and their contribution to the joint number, and judging on the second. Or give one owner the whole path, so that a flow has a single accountable person rather than three functional ones. Each has costs, and picking none of them is the default that keeps the current outcome.

4. Define the joint metric before the technology lands

The discipline here is borrowed and well established. The authors of Trustworthy Online Controlled Experiments, describing practice at Google, LinkedIn and Microsoft across more than 20,000 experiments a year, treat the Overall Evaluation Criterion as the first decision an organization makes: a single metric the whole organization optimizes, agreed before results arrive. Logistics rarely has one. Total cost to serve is the natural candidate, and agreeing it after the orchestration engine is live means agreeing it while three functions are actively unhappy.

5. Set the trade-off policy, not the trade-off

An orchestration engine can weight objectives any way you ask. What it cannot do is decide how much service attainment a point of transport cost is worth, because that is a commercial judgment about your customers and your margins. Write the weighting down as policy, review it quarterly, and keep it out of the daily planning conversation where it becomes a negotiation between functions every morning.

6. Handle cross-company functions as contract terms

Where a degraded number sits inside a carrier or a 3PL, it needs a commercial mechanism rather than an operational one: a gain-share, a repricing trigger, or a volume commitment that compensates for the local inefficiency you are asking them to absorb. Expecting a provider to accept a worse cost per drop because it helps your network is expecting them to fund your optimization.

Three ways to fix the measurement, compared

None of these is free, and the comparison exists because the usual approach, which is to do none of them and rely on goodwill, has a predictable outcome. The functions optimize locally, the orchestration engine is configured to match, and the 4.7% is never collected.

The difficulty also scales with how many functions a flow crosses. A two-function flow needs one concession negotiated; a four-function flow needs three, each with its own owner, its own scorecard and its own view of what is reasonable. That is why orchestration tends to succeed first on short flows inside a single business unit and stall on the long cross-functional ones, which are precisely the flows carrying the largest available gains.

Five questions to ask before an orchestration program starts

What does each function give up, in its own units? If nobody has computed this, the program is asking for concessions whose size is unknown to everyone involved.

Who is measured on the number that improves? The answer determines whether the initiative has an advocate or only opponents.

Which of the three measurement fixes are we using? Naming one is the decision. Leaving it open means the default, which is local optimization.

Who sets the weighting between objectives, and how often is it reviewed? If the answer is the planning team each morning, the weighting is being renegotiated daily by people without the authority to set it.

Which degraded numbers sit outside the company? Those need contract terms, and discovering them mid-deployment is what converts a technology rollout into a commercial renegotiation.

What this looks like in enterprise deployments

A leading North American retailer running multimodal logistics automation across several hundred stores replaced six legacy systems and now achieves 99%-plus on-time store delivery, 95%-plus route compliance, more than 80% reduction in manual dispatch and exceptions resolved in under two hours, with break-even in year one. The six-system consolidation is usually read as an integration achievement, and it is also a measurement one: collapsing six systems collapses the number of places a local optimum can be computed and defended, which makes a joint plan easier to adopt than it would be across a fragmented estate.

A Fortune 50 parcel operation running centralized dispatch across a 120-country network and 51 sites surfaced more than $14 million of unused capacity, including $565,000 at a single site, while lifting weekly execution adherence from 75% to 92%. Unused capacity is precisely the kind of value that no single function is measured on. It sits between sites and between functions, which is why it accumulated, and why finding it required looking at the network rather than at any node’s own numbers.

Four mistakes enterprises make on orchestration incentives

Diagnosing resistance as culture. If the joint plan degrades every scorecard, resistance is the rational response and collaboration workshops will not shift it. The diagnosis determines whether you reach for a measurement change or a change-management budget, and only one of them works.

Assuming integration solves it. A fully integrated stack computing a joint plan produces the same degraded local numbers. Integration is necessary for orchestration and insufficient for adoption.

Leaving the weighting to the planning team. Deciding how much service is worth in transport cost is a commercial judgment. Delegating it by default to whoever configures the engine puts a pricing decision in an operational seat.

Ignoring the functions that sit outside the company. A carrier or 3PL asked to absorb a local degradation has a contract, not a scorecard. That is a commercial conversation and it does not resolve itself operationally.

How Locus makes the trade-off explicit

Locus, the world’s first Decision-Intelligent, Agentic TMS, computes the joint plan rather than a sequence of local ones. Dispatch planning runs against more than 250 real-world operating constraints spanning capacity, time windows, service commitments and cost, and allocation extends across owned fleet, contracted transporters and a network of more than 1,000 carriers, so the plan that gets produced is the network plan rather than transport’s view of it.

The part that matters for this problem is the record. Explainability and Traceability capture the trigger, context, reasoning, action and outcome for each decision, including what was considered and why one option was chosen, which is what turns each function’s concession from a claim into a measured quantity. A warehouse lead arguing that the plan costs them eight percent can be answered with the number, the alternative and the network effect, rather than with an appeal to the bigger picture. Because the route planning system re-optimizes in roughly two minutes, weighting changes can be tested against real volume before they are adopted as policy.

Three boundaries belong here, and they matter more than usual on this topic. Locus does not set the weighting between competing objectives, because how much service attainment is worth in transport cost is a commercial judgment about your customers and your margins. It does not change what your functions are measured on, which is a compensation and reporting decision owned by your leadership. And where a degraded number sits inside a carrier or a 3PL, the resolution is contractual rather than operational, and the platform’s role is to show the size of the effect, not to negotiate it.

Locus supports more than 360 enterprise customers across 30-plus countries, with over 1.5 billion deliveries optimized, more than $320 million in documented client logistics savings and 99.99% uptime. It has been recognized by Gartner for seven consecutive years, featured in the 2026 Hype Cycle for Supply Chain Execution and Logistics Technologies, named a Leader in TMS by QKS Group (SPARK Matrix), and ranked #1 in Route Planning on G2’s 2026 Best Software Awards.

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

So why does logistics orchestration stall inside enterprises that have already integrated their systems? Because the jointly optimal plan is worse than the local optimum for every function it spans, by construction. On a worked case the network improves 4.7%, from 337 to 321, while pick worsens 8%, transport 7% and service 6% against their own bests, and the gain lands on total cost to serve, which typically has no owner with a scorecard. That makes resistance rational rather than cultural, so the remedy is a measurement change rather than a collaboration exercise: move functions onto a joint metric, hold them harmless by judging their contribution to it, or give one owner the whole path. Locus supports this by computing the joint plan across more than 250 real-world operating constraints and recording what was considered and why through Explainability and Traceability, so each function’s concession is a measured number rather than an argument, while the weighting between objectives stays a commercial decision that belongs to you. Request a Locus assessment to quantify what your own joint plan costs each function.

Frequently Asked Questions

Why does orchestration make individual functions’ metrics worse? Because each function’s own best plan was already optimal for its objective, so any plan balancing several objectives is worse for each one individually. On a worked case the joint plan costs pick 8%, transport 7% and service 6% against their own bests while improving the network total by 4.7%.

Is resistance to orchestration a culture problem? Usually not. If the joint plan degrades every scorecard, objecting is the rational response to how people are measured. Treating it as culture leads to collaboration programs that cannot move an incentive structure, which is why those initiatives tend to stall in the same place.

Does better integration fix this? No. A fully integrated stack computing a joint plan produces exactly the same degraded local numbers, because the conflict comes from having several objectives rather than several systems. Integration is necessary for orchestration and insufficient for its adoption.

What are the options for fixing it? Three work. Move functions onto a joint metric such as total cost to serve; hold them harmless by reporting their local number but judging their contribution to the joint one; or give a single owner an entire flow so accountability follows the path rather than the function. Choosing none of them is the default that preserves local optimization.

Who should set the weighting between competing objectives? Leadership, as written policy reviewed on a fixed cadence. How much service attainment is worth in transport cost is a commercial judgment about customers and margins, and leaving it to whoever configures the engine puts a pricing decision in an operational seat.

What happens when a degraded function sits inside a carrier or 3PL? It becomes a contract question. A provider asked to accept a worse cost per drop for your network benefit is being asked to fund your optimization, so the mechanism has to be commercial: a gain-share, a repricing trigger, or a volume commitment that compensates for the inefficiency.

How do we start if nothing has been quantified yet? Compute each function’s concession under the joint plan in its own units and put those numbers on one page beside the network gain. That single page changes the conversation from a request for goodwill into a negotiation with known quantities, which is the only version that resolves.