Lead Time: Why Your Estimate Is Always Too Low

The symptom

"A standard order takes us about eight days." I hear that sentence in every second initial conversation, usually delivered with confidence and confirmed by everyone in the room.

Measured, the number is almost always substantially higher — not by ten or twenty per cent, but frequently by a factor of two. And this isn't a question of honesty. People know their own work very precisely. What they don't know is the waiting time in between, because waiting time belongs to nobody.

The mechanism

An order's lead time consists of two entirely different components: processing time and queue time. Processing time is what somebody actually does. Queue time is what happens while nothing happens.

Asked about lead time, people add up processing times in their head. Two hours of work preparation, one day of production, half a day of assembly, half a day of dispatch — call it three days plus buffer, say eight. That calculation is correct for every individual step. It simply omits that the order sits between steps, and sits considerably longer than it is worked on.

In the processes I have measured, processing time as a share of total lead time is often in the low single-digit percentages. The rest is waiting: for an approval, for a machine, for a part, for an answer to a query, for Monday.

Why does nobody notice? Because queue time lands in no confirmation anywhere. Every participant sees the order at the moment they work on it, and in that moment they are fast. Between those moments, nobody sees it at all. Queue time is the sum of the periods during which the order isn't on a desk but in a queue — and queues appear in no system.

There's a second effect: people remember the normal case and deliver the mean. But the mean includes the two orders a month that took four weeks because a query to the customer was outstanding. Scheduling from experience means scheduling to the median, then being surprised by the mean.

The cost

An anonymised example, altered in detail, real in pattern. A manufacturer of special-purpose machinery routinely quoted customers a six-week delivery time. Internally that figure was regarded as conservative.

We measured 74 completed orders over twelve months, with timestamps at seven handover points. The median was 41 calendar days — just inside the commitment. The mean was 58 days. The upper quartile was over 70.

Pure processing time across all steps summed to about 4.5 working days. Everything else was queue time. The largest single item: an average of 9.3 days between technical clarification and production release, mostly due to open queries with the customer that nobody was systematically tracking.

The cost didn't sit in production, it sat in sales. 23 per cent of orders exceeded the promised delivery time. Four led to contractual penalties in the period examined, several to discounts on follow-on orders. And the business had started building safety buffers into its commitments — which lost orders to faster competitors, even though the real processing time was entirely competitive.

The fix

First, measuring replaced estimating. Seven timestamps along the process, with no new system — the points already existed, they had simply never been analysed.

The main item identified was the customer query loop. It was defused with two simple rules: queries are bundled and raised within 24 hours of technical review, not drip-fed. And every open query gets a date and an owner; after three days without an answer, somebody picks up the phone instead of waiting.

Second, the delivery commitment moved from experience to the distribution. What gets promised isn't the median but a figure that 90 per cent of orders meet. Initially that was an uncomfortable lengthening of the commitment — until it became visible that seven weeks kept is worth more than six weeks missed.

The mean fell to 46 days in the following year and the overrun rate to under 5 per cent. The commitment could then be shortened again, this time with cover behind it.

This separation of processing from waiting is the core of a process analysis: measuring an order's path end to end with real timestamps, rather than adding up remembered durations. Fixed scope, fixed duration, fixed price.

The next step

Take your last twenty completed orders and count calendar days from order entry to dispatch. Compare the mean against your commitment. You don't need a project for this — one hour and your own data.