Most approval processes in established companies were not designed; they accumulated. Each additional signature can be traced to a specific incident, and almost none of them were ever removed afterwards.
Adding a reviewer is the cheapest visible response
When something goes wrong, a manager needs to show a correction. Investigating the underlying cause is slow and uncertain, while adding a review step is immediate and easy to describe upward.
The new step also transfers risk. Whoever now signs shares responsibility for the next failure, which makes the addition attractive to the person proposing it.
Nothing in the process measures what the step costs. The delay it creates is distributed across everyone who submits work, and none of them own the decision to keep it.
Removal has no natural sponsor
Someone who deletes an approval step owns any failure that follows, and receives nothing if none does. The incentive is entirely one-sided, so the steps stay.
The reviewers themselves rarely object either. Sitting in an approval chain is a form of visibility, and being removed from one reads as a loss of standing.
So the chain lengthens monotonically. A ten-year-old company can accumulate approval paths nobody living remembers designing.
Long chains change what people submit
Employees learn how long approval takes and adjust their requests to avoid it. Work gets split below thresholds, bundled into fewer larger requests, or routed around informally.
Splitting is the most common. A request just under a review limit moves quickly, so budgets are spent in pieces that individually escape scrutiny.
The result is that the control designed to increase oversight decreases it, because the traffic it was meant to inspect has reorganised itself to pass underneath.
Reviewers stop reading once volume rises
An approver handling a handful of items a week reads them. One handling dozens a day develops heuristics, and eventually approves on the basis of who submitted rather than what was submitted.
This is not negligence so much as arithmetic. Careful review takes time the approver's calendar does not contain, and the queue does not shrink to accommodate it.
At that point the step still costs delay while providing very little inspection. It has become a queue rather than a control.
Pruning requires a rule, not a judgement
Companies that successfully shorten chains generally do it by rule rather than case by case, because case-by-case removal puts the risk on a named individual each time.
Common rules include expiring any approval step after a fixed period unless renewed, or requiring that every new step replace an existing one rather than add to it.
Both work because they move the burden of justification onto keeping a step, which is the opposite of how the chain grew in the first place.