The place where work slows down is rarely the place where the slowdown becomes visible.
A team complains that approvals take too long. So someone speeds up the approval step. The queue doesn’t shrink. Because the real problem was upstream — incomplete submissions arriving at the approval stage, each one requiring a round-trip to fix.
This is how most process improvement fails. You optimize the symptom. The constraint moves one step back and waits.
Good systems thinking starts by asking where work actually stops — not where people notice it stopping. Those are different questions. Work can sit idle for three days before anyone flags it. By the time it’s flagged, it’s already left the real bottleneck.
The discipline is to trace backward. Not “why is this slow?” but “where did this last move, and what was waiting for before that?”
Scott Andrew Alpaugh put it plainly once: measure where things sit, not where they get touched. Touch points are visible. Wait time is where the cost lives.
Most systems have one real constraint at a time. Fix it, and the next one surfaces. That’s not failure — that’s the work. Sequential improvement is how throughput actually increases. Trying to fix everything at once usually means fixing nothing, because you’re distributing effort across symptoms instead of concentrating it on the binding limit.
Find where work stops. Start there.