Bottlenecks: Finding the Leverage Points

A busy team can improve many individual tasks without improving the result that matters. If finished work is waiting for approval, producing more work upstream may only lengthen the queue. Start by finding what limits the system, then check whether your measures and incentives help you address it.

1. False Proxies

A proxy stands in for an outcome that is hard to measure directly. It becomes misleading when people can improve the metric without improving the outcome. Hours worked can reward being present rather than useful work; pressure on customers can raise satisfaction scores without improving their experience.

Campbell’s Law describes the pressure to corrupt a social indicator when decisions depend on it. Goodhart's Law is commonly expressed as: when a measure becomes a target, it ceases to be a good measure. Ask how someone could raise the number while making the underlying result worse.

2. Theory of Constraints (TOC)

Eliyahu M. Goldratt’s approach focuses improvement on the constraint that limits throughput. Its focusing steps are:

1. Identify the constraint.
2. Exploit it: use the available capacity more effectively.
3. Subordinate other work to it: coordinate the system around the constraint.
4. Elevate it: add capacity or make a larger change if needed.
5. Repeat when the constraint moves.

TOC evaluates changes through throughput, investment and operating expense. Making a nonconstrained activity faster may do little for overall throughput.

3. Diminishing Returns

Increasing one input while holding others constant can eventually produce smaller additional gains. Excess fertilizer can harm crops; extra study time can displace sleep. Adding people to a project also adds coordination costs, so it is worth checking what the extra capacity will actually accomplish.

4. Pareto Principle (80/20 Rule)

The heuristic that roughly 80% of consequences come from 20% of causes directs attention to uneven contributions. The actual shares must be measured; the rule does not establish a power-law distribution. Possible examples include 80% of profits coming from 20% of customers, or 80% of crashes from 20% of bugs. Use the idea to look for concentration, rather than forcing observations to fit it.

5. Via Negativa

Before adding a process, ask whether removing an obstacle would help. Cutting unnecessary approval steps or deleting confusing prose can be useful. Removal also has risks: a control may look unproductive until it prevents a failure. Check what the proposed subtraction does and what depends on it.

6. Regulatory Capture

A regulator can come to serve the interests of the industry it is meant to oversee. Dependence on industry information, lobbying and career incentives can contribute. Examine who influences the rules and whose interests are excluded; a particular regulatory arrangement alone does not establish capture.

7. Boiling Frog Syndrome

The boiling-frog metaphor describes failure to respond to gradual deterioration. Track changes against an explicit baseline so that a succession of small departures does not disappear into a new sense of normal. The metaphor is useful for discussing scope creep, declining quality or a deteriorating relationship.

8. Normalization of Deviance

Diane Vaughan’s account of the Challenger disaster examines how repeated departures from a standard can become accepted when they do not immediately produce catastrophe. A launch that survives does not by itself establish that the accepted risk was sound. The same question applies when teams repeatedly bypass a safety check or a trader takes larger risks after earlier bets pay off.

9. Cobra Effect

An incentive can make the problem it addresses worse. The name comes from a story about a colonial cobra bounty that encouraged breeding snakes; ending the bounty then encouraged their release. The general question is how people might respond to the reward in ways the designer has overlooked. Rewarding closed support tickets, for example, can encourage staff to close unresolved cases.

Applying the Models

1. Find the constraint that limits the result you want.
2. Check whether the metrics reveal that result or reward a substitute.
3. Look for an obstacle you can remove, and check what its removal might damage.
4. Concentrate effort where the evidence suggests it will matter most.
5. Review repeated exceptions and gradual departures from standards.
6. Test how people could game a new incentive. Where regulation matters, examine who shapes the rules.

After a change, measure its effect on the system. A successful improvement can move the constraint elsewhere.