Look beyond the official process map
Bottlenecks often live between systems and teams rather than inside a documented step. Observe queue age, handoffs, duplicate entry, approval wait, exception paths, and the information people recreate to make a decision.
Combine interviews with actual artifacts such as forms, reports, tickets, messages, and timestamps. Perception identifies where to look; evidence shows what is happening.
Distinguish symptoms from constraints
A backlog may reflect insufficient capacity, but it can also result from incomplete inputs, unclear priority, batch processing, or rework upstream.
- Where does work wait longest?
- Which inputs are frequently missing?
- Which decisions depend on one person?
- Where is data copied or reconciled?
- Which exceptions repeat often enough to design for?
Prioritize the smallest useful intervention
Do not redesign an entire operating model when a better intake check or routing rule can remove the constraint. Rank interventions by likely impact, evidence confidence, effort, risk, and time to measurable learning.
Re-measure after the change. Removing one bottleneck may expose the next, which is a normal part of continuous improvement.