Improving flow: finding where work stops and deciding what to attack first

Touch time is a small share of elapsed time, and the gap is the prize

Take one representative order and record when each operation started and finished, including the waiting in between. The sum of the operations is normally a small fraction of the elapsed time from release to completion. That arithmetic reframes the conversation, since speeding up an operation by a worthwhile margin barely moves the elapsed time, whereas removing one place where work sits for two shifts moves it a great deal. It also identifies which departments own the delay, and in most plants a meaningful portion of the waiting happens outside production altogether.

Choose one family and walk it, physically

Flow is a property of a path, and a plant has many paths. Trying to improve all of them together produces generalities. Pick one product family with enough volume to matter, walk the route in the direction the material travels, and stand at each stopping point long enough to discover why work waits there. The answer is often something nobody would have written on a chart, such as a single trolley shared between two areas or an inspector covering three departments. Documentation gathered in a meeting room describes the routing, not the process.

Stability, then flow, then pull

The order of intervention matters. Where equipment stops unpredictably, material arrives short or quality is erratic, connecting operations more tightly makes the plant worse, because disturbances now propagate instead of being absorbed. Fix the sources of disruption along the chosen route first, even though it feels like a detour. Then reduce the batch and the distance between steps so work moves continuously. Only after that link the steps with a signal limiting what may be released. Plants that install the signalling mechanism first get a system overridden within weeks and a floor that has learned to be sceptical.

Improvements that move the queue somewhere else

A faster operation, a new machine or a rebuilt cell often shifts the accumulation rather than removing it, so work now waits before the next step, in a staging area that did not previously exist, or in the despatch bay. That is not automatically a failure, since moving a queue somewhere more visible or less costly can be worthwhile, but it should be recognised rather than reported as a lead-time gain. The check is to measure the whole route before and after, count the work in progress on it, and confirm that finished items actually left the building sooner.

Somebody has to own the whole path

Departments optimise what they are measured on, and flow runs across them. Without a named person accountable for the elapsed time of a product family from end to end, holding the authority to ask a department to run smaller batches for the greater good, improvement stalls at the first boundary. That role is uncomfortable because it carries responsibility without direct control, and it needs visible backing from the site leadership. Where it does not exist, expect flow projects to succeed inside single departments and to die at the handovers between them.

Frequently asked questions

Where should we start if the whole plant flows badly?

With a single product family that matters commercially and whose route can be walked in an afternoon. Broad diagnosis across every product yields a picture too general to act on and consumes the goodwill needed for the first change. One route, walked and measured properly, gives a concrete list ranked by how much waiting each stoppage causes, and delivers a visible result that makes funding the second route considerably easier.

How do we improve flow when several products share the same machines?

Accept that the shared resource is a scheduling decision rather than a flow problem you can design away. The options are dedicating equipment where volume justifies it, shortening changeovers so the machine serves more families more often, or holding a controlled buffer at that step and treating it as designed inventory. The mistake is designing cells that quietly assume exclusive access to a machine which three other cells also depend on.

Does better flow always reduce cost?

It reduces working capital, scrap discovered late, expediting and the administrative effort of chasing, and it usually improves delivery. It can raise unit cost measures that reward long runs and full utilisation, because smaller batches mean more changeovers and some equipment will stand idle. Whether the trade is worth taking depends on the business, and it must be settled before the change, since otherwise the standard cost report will be used to reverse it.