Finding bottlenecks: measure waiting time, don't guess

The lead time of an order is the span between the moment it arrives in the company and the moment it is completed. It is the figure the customer experiences. Handling time, by contrast, is the sum of the minutes in which someone is actually working on that order: typing, checking, assembling, picking, signing. Between the two figures there is a factor in nearly every company that surprises everyone involved when it first appears on the table. An order that is in the building for twelve calendar days often adds up to three to five hours of genuine handling. The difference is not sluggish staff, it is a property of the workflow.

This distinction has an uncomfortable consequence: anyone who wants to become faster and starts with handling time is working on the smaller part of the problem. If eighty per cent of lead time is waiting (project experience), even halving the actual working time shortens the flow by only a few per cent. That explains why companies invest in faster tools, new screens or an additional member of staff and then find that nothing has changed in the delivery time promised to customers. Time was saved in a place where it was never lost.

A bottleneck assessment rarely fails on method and often on the fear of effort. Yet it needs to be neither a project with a steering committee nor surveillance of employees. What is required are two timestamps per station: the moment the case arrives at that station and the moment work on it begins there. The difference is the waiting time. A third stamp for the end of processing yields the handling time. The first round needs nothing more. It is important that the stamps belong to the case and not to the person — this is about the path of the order, not about individual performance. Say so before the measurement starts, otherwise mistrust makes the figures useless.

The scope can stay small. Twenty to thirty orders over three weeks are enough for a sound picture, provided they cover the normal range and not just the rush jobs. A single day is not enough, because weekdays, holiday periods and month-end closing change the flow. If the assessment can be tied to ongoing metrics and reporting because the timestamps already exist in a system, do that. Where this is not possible, a routing slip or a shared list works just as well — the insight comes from separating waiting and working, not from the tool.

Approvals are the case in which handling time and waiting time diverge the most. A signature or a tick takes barely a minute, but the case waits for one particular person to reach a particular place or a particular program. If the approval comes in the evening, the case has lost a working day without anyone being idle. If it falls on a Friday afternoon, that becomes three calendar days. This is why a chain of three approvals can create a week of lead time even though five minutes of work were done in total.

Before technology comes the question of necessity: which approval protects against a real risk, and which exists only because it was introduced at some point? Value thresholds set ten years ago often catch routine cases today. Where an approval is to be kept, a clear deputy rule usually helps more than any software. Only then is it worth looking at a digitally mapped approval as implemented within process automation — with notification, deputies and a deadline after which the case is escalated.

The most important expectation to set before any improvement: a bottleneck that has been removed does not disappear, it moves. As soon as the approval no longer costs a day, work arrives earlier and in larger volume at the next station — and that is where the new queue forms. This is not a sign of a failed project but the normal course of events. Every bottleneck you remove makes the next one visible, and that second bottleneck was previously hidden underneath the waiting time of the first.

From this follows a practical rule: change one thing at a time and measure again afterwards with the same measuring points. Anyone who implements three measures at once cannot say afterwards which one worked, and risks leaving an ineffective change in the workflow permanently. The second round of measurement is considerably cheaper than the first, because the stations and the recording method are already in place. In practice ten to fifteen cases are often enough.

Equally important is the question of what happens to the time gained. Shorter lead times only pay off if they turn into something: faster quotations and therefore higher win rates, earlier invoicing and therefore better cash flow, shorter commitments to customers. If the freed-up time is not named, it disperses unnoticed and the improvement cannot be demonstrated later. Around half of all companies cannot fill open positions over the longer term (DIHK) — against that background, the recovered time of experienced staff is usually the most valuable part of the result.

A bottleneck assessment that ends up in a drawer was wasted time. For the insight to last, the new workflow should be written down — briefly, with stations, responsibilities, deadlines and deputies on one or two pages. This short process documentation is also the basis for onboarding and for the next round of measurement, because it defines what is actually being measured. Extensive manuals are not needed for this and are rarely maintained in mid-size companies anyway.

Two side effects should be planned for. First: recording timestamps per case creates data from which conclusions about individual employees may be drawn. That touches data protection and, where a works council exists, its codetermination rights for systems capable of recording behaviour or performance. Clarify purpose, scope and retention period beforehand and limit the evaluation to stations rather than people; a professional review of the individual case remains advisable. Second: a measurement changes behaviour while it runs. Expect the first week to look slightly better than normal operation, and therefore evaluate the whole period.

If you want to start small, pick a single case type with high volume — the standard order, the recurring service job, the incoming invoice. Such a section can be recorded in a few days, delivers a figure nobody had before, and puts the discussion about priorities on a factual footing. Everything else follows from the result rather than from an assumption about where things ought to be stuck.