A function can hit its target and leave the organization worse off. Sales accepts orders faster, but fulfillment cannot process them. A service desk closes contacts sooner, but customers have to call again. Production releases more work, but the next stage has no capacity to handle it. Each team may have improved the result it was asked to deliver. The enterprise still has to absorb what happened next.
Judge an efficiency gain by what happens after the work leaves the team claiming it. A local improvement can damage the whole when its measure excludes the delay, constraint or cost transferred elsewhere. The gain matters only if it survives the journey to the customer outcome.
Where the measure stops
Every performance measure has an endpoint. Time per contact stops when the contact closes. Workstation output stops when work leaves the station. Neither necessarily captures a repeat call, a queue at the next stage or a defect discovered later.
That does not make local measures useless. Teams need timely feedback and room to improve the work they control. The problem begins when a measure of one stage stands in for the result of the entire process. A function can then record a genuine gain while another team inherits the unfinished work. If leaders see only the first report, they may reward a transfer as though it were a reduction.
Donella H. Meadows, author of Thinking in Systems: A Primer, warned against maximizing a part while disregarding the whole. A system boundary can leave consequential outputs out of view. For leaders, the practical task is to examine an improvement across enough of the process to see where its costs and benefits end up.
How a reasonable decision creates friction
Consider an illustrative order process. Sales is measured on how quickly it accepts orders and removes an eligibility check that slows approval. Acceptance time falls. Customers receive confirmations sooner. The removed check, however, also identified orders requiring specialist configuration. Those orders now reach fulfillment without the information needed to complete them.
Fulfillment seeks clarification. Specialists face a longer queue. Some orders return to sales; others proceed and need correction. Sales has improved the speed it controls, but the business may have exchanged a visible delay before acceptance for less visible delays afterward. In this hypothetical scenario, end-to-end order time, rework and customer outcomes would reveal whether removing the check helped.
The same mechanism can take several forms. Work arrives at a constrained team faster than it can be completed. A quality decision moves downstream, where correction takes more effort. One budget falls while another absorbs the expense. A task is marked complete even though the receiving team cannot act on it. In each case, the local gain can be real and still be an incomplete account of performance.
The Goal: A Process of Ongoing Improvement, by Eliyahu M. Goldratt and Jeff Cox, brings a constraint-centered perspective to the flow of work. H. William Dettmer's account of the Theory of Constraints makes the performance test explicit: judge local decisions against system throughput, inventory and operating expense. In the order example, first ask whether faster release improves completion through the limiting stage or simply adds to the work waiting there.
In this case, Strategic Friction appears as a mismatch between the desired outcome and the behavior a measure rewards. Execution Drag describes the waiting, clarification, rework and coordination that may follow. The queue reveals where to investigate; it does not establish that a misaligned measure caused it.
A boundary visible in the evidence
A study by Tim Tenbensel and colleagues offers a qualified example. In four New Zealand hospitals operating under a six-hour emergency-department target, reported department length of stay continued to fall, while a broader measure that included time in short-stay units did not fall further after 2010. Greater use of short-stay units largely explained the divergence.
The study does not show that transfers to those units were inappropriate or that the target brought no benefit. It also reports earlier reductions in total time. Its narrower lesson is valuable: when two measures ended at different points in a patient’s journey, they gave different accounts of later improvement. An executive looking only at the department measure could miss time still spent in the wider care pathway.
An operational handoff deserves the same scrutiny, though the hospital finding cannot be generalized to every industry or target. When a team’s definition of “finished” differs from what the next team or customer experiences, compare the measures. A divergence calls for investigation; it does not, by itself, prove manipulation or failure.
Test the boundary before changing the target
The proposed OutcomesLab System Boundary Test examines a proposed functional improvement. Before approving the change, and again after implementation, ask:
- What exactly improves? Specify the local measure and where its observation stops.
- What happens to the work next? Identify the receiving team, queue, decision or customer outside that boundary.
- What limits completion? Determine whether the change protects the limiting stage, improves its inputs, reduces its load or adds to its queue.
- What would count as an end-to-end gain? Select a completion or customer-outcome measure, with a safeguard for quality, risk or cost.
- Who will review the whole result? Assign an owner able to compare the local gain with downstream effects and act on what the comparison shows.
Apply the test to a real flow—an order, a request or a strategic initiative—not just an organizational chart. Follow it until the intended result is delivered. A faster handoff may be useful, but it is not evidence of a faster outcome. The exercise may also reveal why a seemingly inefficient step exists: an early check can protect a scarce specialist from incomplete work.
Keep the answers in a Friction Transfer Ledger, the working record for this test: who gains, who absorbs work or risk, and what evidence would show an end-to-end improvement. Give the review an owner and date so the comparison leads to a decision rather than another report.
The organizational-learning perspective in Peter M. Senge's The Fifth Discipline: The Art and Practice of the Learning Organization matters here too. Teams must be able to report an uncomfortable result: their metric improved while the overall outcome did not. Leaders need that information to change the design of work. Blaming the function that reports it would make the next problem harder to see.
Keep local accountability—add wider feedback
Local targets still have an essential role. They provide prompt feedback, clarify responsibility and let specialists improve work they understand. Where interfaces are clear, capacity is adequate and measures reveal the consequences of handoffs, local improvement can support the whole. Meadows also cautioned against excessive central control; widening the evidence should not require moving every decision to the executive team.
Evidence also argues against treating a tight local target as inherently harmful. In a retrospective study of 15 purposively sampled English emergency departments, Ellen J. Weber and colleagues found no apparent worsening in the quality and safety outcomes they measured under a four-hour target, and little effect on test use. Those selected measures cannot rule out every unintended consequence. They do rule out presenting damage as the inevitable result of a time target.
The choice, then, is not between local accountability and systems thinking. Keep measures that help teams act. Add feedback from the wider outcome when consequences cross boundaries, and investigate when local and end-to-end results move in opposite directions. Do not remove a quality or safety control merely because it slows one stage; first establish what work that control prevents downstream.
Executives can begin with an outcome that is missing its mark while contributing functions report healthy performance. Trace one flow from request to result. Compare each team’s definition of completion with what the next team actually receives. Look for growing queues, repeated clarifications, corrections and costs that have changed owners rather than disappeared.
Use the System Boundary Test on the next proposed efficiency gain. If performance improves only before the handoff, revisit the change or its measure. If both the local result and the customer outcome improve without weakening safeguards, keep the change and learn why it worked. Reserve the claim of success until the organization has delivered the result.