CompanyGraph
How missing signals, delayed information, and blocked corrective action let small operating problems become structural decline.
A Feedback Loop Needs an Action at the End
A business learns through a chain: an operating condition changes, someone observes it, information reaches a decision-maker, a response is authorised, and the result is measured again. Feedback is not merely negative sentiment or a KPI. It is a connection between a condition and an action that can change it.
The loop can fail in several places. A sensor may not measure the relevant condition. A report may arrive after inventory, cash, or customer trust is gone. A summary may remove the location or product that needs attention. A manager may see the signal but lack budget, authority, or time to respond.
Four Failure Modes
Blindness occurs when the system does not measure the relevant condition: a company tracks shipments but not customer usability, or revenue but not renewal quality.
Delay occurs when the signal arrives after the next production, hiring, or financing decision. Long contracts and annual reporting can hide deterioration for months.
Filtering occurs when bad news is softened or removed as it passes through layers. Aggregation can be useful for scale while hiding the store, customer, or machine where the failure is occurring.
Blocked action occurs when the signal is accurate but the responsible participant cannot change the process. A supplier may report a defect while the buyer’s contract forbids a specification change; an engineer may identify a safety issue while a budget holder delays maintenance.
The Metric Is Not the Condition
Customer-satisfaction scores can rise while the most profitable customers leave. On-time delivery can improve while the wrong products arrive. A risk dashboard can show low exposure because a position is outside the measured entity. Each metric is useful within its definition; none is the physical condition itself.
Good feedback design keeps enough identity, time, and location to connect a signal to a decision. It also records what happened after the response. Without that second observation, management cannot know whether the action corrected the cause or merely moved the problem.
Silicon Valley Bank Shows a Broken Response Path
The Federal Reserve’s review of Silicon Valley Bank describes warning signals involving interest-rate risk, concentrated funding, liquidity, and supervision. The case is not proof that all risk dashboards fail. It demonstrates a specific path in which recognised concerns did not produce timely corrective action before withdrawals accelerated. The Federal Reserve review supports the distinction between detecting a risk and acting on it.
Money and Incentives Shape the Loop
Measuring a problem costs money. Correcting it can cost more: a recall, a second supplier, extra staff, a slower line, or a cancelled project. If performance pay rewards shipment volume while defects appear later, the person closest to the signal may not be rewarded for reporting it. If the budget arrives after the maintenance window, the corrective option may have closed.
Leadership can improve the loop by assigning ownership, protecting escalation, funding tests and redundancy, and reviewing failures without confusing a write-off with incompetence. Those controls change the economics of truthful feedback; they do not guarantee that every signal is correct.
How to Audit a Feedback Loop
- What condition is being sensed, at what resolution, and with what delay?
- Who receives the signal, and what information is lost in aggregation?
- Who can authorise the response, and what money, equipment, or time is required?
- How is the result measured after the response?
- What happens when the signal contradicts the target used to reward the team?
A company is learning when observations can still change the work that produced them. When the chain from condition to correction is delayed, filtered, or unfunded, the organization may report control while operating without it.