Before you speed anything up, find out where the time actually goes — it is almost never in the work.
Map the process before you change it. Draw every step from order to delivery, mark which steps actually add value the customer would pay for (VA) and which don’t (NVA), and measure two clocks: (order in to order out) and touch time (the real work).
The gap between them is where the waste hides — usually as waiting. In most processes the item is worked on for minutes and waits for hours. Baseline that honestly and improvement becomes measurable instead of a hunch.
a value stream for one order — tag the steps, shrink the batch
Two clocks and one ratio. Baseline them for the current state, then attack the biggest block of time — which is almost always the waiting between steps, not the steps themselves.
lead time = clock time, order in to order out
touch time = actual work time = 40 min (constant — the work doesn't change)
value-adding (pick + pack) = 18 min
non-value (log + approve + inspect) = 22 min
waiting (batch of 10): each order waits while the rest of its batch is
processed at every station
~ (batch - 1) x touch = 9 x 40 = 360 min
lead time = touch + waiting = 40 + 360 = 400 min
cycle efficiency = touch / lead = 40 / 400 = 10%
switch to one-piece flow (batch = 1):
waiting -> 0 lead = 40 min efficiency = 100%
same work, delivered ~10x sooner — you deleted the waiting, not the work.
Notice what didn’t move: the 40 minutes of touch time. Speeding up the 10-minute pick step would have shaved a rounding error. Shrinking the batch cut the lead time tenfold.
The eight wastes (mnemonic DOWNTIME) — what to look for on the map:
| Value-stream mapping fits when… | The trade-off |
|---|---|
| A repeatable process has hand-offs, queues, and a clear start and end. | Needs honest current-state data — measured, not remembered. |
| Lead time hurts the customer and no one can say where it goes. | Maps a single product family well; high-mix flows need one map each. |
| You want a baseline so improvements are measurable. | Less useful for one-off, creative, or highly variable work. |
An interviewer says: “Order-to-ship takes us five days, but there’s maybe 40 minutes of actual work in it. Where do you look?” A strong answer resists the urge to optimise a step. First, map the current state and measure lead time versus touch time — a cycle efficiency near 1% tells you the five days is almost entirely waiting. Then find the biggest queues: orders batched before shipping, an approval sitting in someone’s inbox, inventory stacked between stations. Attack those — smaller batches, one-piece flow, pulling approvals inline — and re-baseline to prove the lead time dropped. You explicitly do not start by buying a faster picking machine, because the picking was never where the time went.
Check yourself
Lead time is five days; touch time is 40 minutes. Where is the biggest opportunity?
You cut the batch from 10 to 1. Touch time is unchanged at 40 minutes. What happens to lead time?