Common Improvement Traps

Brian Joiner identified three levels at which any problem can be fixed: the output (deal with what came out wrong), the process (change how the work is done), or the system (change the conditions that make the process produce poor outputs). Most organisations operate at Level 1 — fixing outputs — because it is immediate, visible, and rewarded. Level 1 fixes are also the ones that produce recurrence: the system is unchanged, so it continues to produce the same outputs.

The trap closes when Level 1 becomes the habitual response. Every new problem gets a new fix at the output level. The backlog of structural problems grows while the organisation is busy with symptoms.

When organisations are measured by audit results, they allocate effort to producing good audit results. Documentation, training records, and audit preparation become the primary outputs of quality and safety work. The audit measures the organisation’s ability to appear compliant, not its actual safety or effectiveness. The gap between the two can be very wide — and is invisible to the compliance system, because the compliance system cannot see what it is not looking for.

The trap closes when compliance becomes the ceiling rather than the floor: when the question “did we pass the audit?” replaces the question “is the system actually producing safe outcomes?”

Every process produces variation. Some of that variation is common cause — the normal fluctuation of a stable system — and some is special cause — a genuine signal that something structural has changed. Tampering is what happens when a manager reacts to common cause variation as if it were a special cause signal: adjusting the process, adding a correction, issuing a reminder, in response to what is actually random noise.

The result is a system that is worse than if it had been left alone: every tampering response adds a new source of variation, making the system less predictable over time. W. Edwards Deming demonstrated this with the funnel experiment: adjusting the funnel to compensate for each previous result produces a pattern far more scattered than simply leaving it fixed.

John Seddon identified two types of demand on any system: value demand (what the customer actually wants) and failure demand (demand caused by a failure to do something right first time). A patient who calls back because their prescription was wrong is failure demand. A customer who contacts support because the online process failed is failure demand. Failure demand typically makes up 30–70% of all demand in service organisations — and is largely invisible, because systems are designed to handle calls, not to classify the reason for them.

The trap is to respond to rising demand by adding capacity — more staff, longer hours, faster processing — rather than eliminating the failure that creates the demand. Adding capacity to handle failure demand is a Level 1 fix: it manages the symptom while the structural root continues to generate new demand.

Clayton Christensen documented how successful organisations are systematically destroyed by disruptive innovation — not because they make mistakes, but because they make entirely rational decisions that are correct for the current business model and fatal for the future one. They listen to existing customers, invest in existing products, and allocate resource to existing markets. Each decision is individually defensible. Collectively they make it impossible to respond to a challenger who is approaching the market from a different direction.

In improvement terms, the trap appears when protecting existing performance prevents adopting the structural changes needed for future performance. The organisation cannot change because the current system is working — and change would disrupt the current system.

Eliyahu Goldratt’s “necessary but not sufficient” framework identifies a trap that is almost universal in complex improvement programmes: a solution is identified, implemented, and partially delivered — and the expected result does not appear. The solution was necessary. It was just not sufficient. Other conditions, not addressed by the same intervention, are also required. Technology alone is rarely sufficient. Training alone is rarely sufficient. A new process alone is rarely sufficient.

The trap closes when the partial result is interpreted as evidence that the solution does not work, rather than evidence that additional necessary conditions remain unmet. The organisation abandons the improvement and moves to the next initiative, without ever having completed the solution.

The Theory of Constraints establishes that every system has one dominant constraint: the single bottleneck that limits overall throughput. Improving anything that is not the constraint produces no improvement in system output. It may produce local efficiency gains — a process that runs faster, a team that is less busy — but if the constraint is elsewhere, the overall performance stays the same.

The trap is to improve what is visible, measurable, and easily changed rather than what is actually limiting output. The constraint is often an unscheduled human decision, a missing system, or a structural absence — things that do not appear on process maps. Improving the processes around the constraint does not help.

Ashby’s Law of Requisite Variety states that only variety can absorb variety: a control system can only manage the variety in its environment if it has at least as much internal variety available to respond. The trap appears when a low-variety response — a checklist, a policy, a compliance framework, a standardised process — is applied to a high-variety problem. The response absorbs some of the variety and passes the rest to the nearest available expert human, who becomes a bottleneck.

This trap is closely linked to the Compliance Trap and the Wrong-Level Fix Trap. The variety mismatch is the structural reason why compliance systems cannot prevent what they are designed to prevent: they have insufficient internal variety to respond to novel failure modes.

A system is a set of components that work together to produce an outcome. Optimising individual components independently does not optimise the system — it often harms it. Deming called this “sub-optimisation”: when a department, team, or function improves its own performance metrics without regard to the effect on adjacent parts of the system, the overall system performance may decline even as all local metrics improve.

The silo trap is built into the standard management structure: functions are accountable for functional metrics, not for the end-to-end outcome the system is designed to produce. A procurement team that achieves its cost-reduction target by switching suppliers may increase rework rates in production. A discharge team that meets its bed-turnaround target by discharging patients earlier may increase readmission rates. Each part looks better. The whole is worse. See Systems Thinking & Silos for the full treatment.

The PICK model identifies the Scatter Trap precisely: organisations that pursue too many improvement initiatives simultaneously make slow progress on all of them. Each initiative consumes some management attention, some implementation capacity, and some staff time. No initiative receives enough of any resource to reach the point of structural change. The organisation is always busy and never changing.

The scatter trap is reinforced by the political and managerial incentives around initiative launches. Starting an initiative is visible, credible, and rewarded. Stopping an initiative in progress is painful. Organisations accumulate active initiatives over time, each drawing on the same constrained pool of implementation capacity, none of them receiving what they need to succeed.