The target that was met twice

The target

Kemal Gubran read Marta Ocklund's letter and did what a managing director does. He gave Nadira a number.

On-time delivery to Palliser, on Palliser's own definition, at or above 95% by the end of the following quarter. Baseline 71.4%. Reviewed monthly. It was a good target. It used the buyer's definition instead of the whiteboard's, it named a date, and it was written down.

Nadira took the next quarter's Palliser ledger: 91 delivery lines due, 65 shipped complete on or before the confirmed date, 71.4% again. Then she modelled two ways of reaching 95%.

Two moves, one number

Move A: quote the date from the plan. For years, Velanthur's order-acceptance desk had confirmed whatever date the buyer asked for. Saying yes wins the order, and the factory would "find a way". The move is to confirm from the actual line plan instead, with the real queue of work in front of it. Applied to the quarter's 91 lines, the confirmed dates move out by 6.4 days on average, and 22 of the 26 failures now land on or before the new date. 87 of 91: 95.6%.

Move B: refuse the short-notice drops. Palliser places top-up drops with very little notice: a colour that sold, a size that ran out. 23 of the 26 failures were lines accepted inside a 45-day window. Refuse those and the quarter has 68 lines due, of which 65 land. 65 of 68: 95.6%.

Two different actions, taken by different departments, with different consequences for the buyer. They land on the same figure to one decimal place. Neither is a trick, and neither involves recording anything differently.

Which one is right depends on a number nobody put in the target

Velanthur runs 14 sewing lines. In the quarter Kemal set the target, the book was full: 1,092 booked line-days against a capacity of 1,120 — 97.5%. A line-day is one sewing line for one working day, and it is the unit a factory sells.

In that quarter, Move A is simply telling the truth. A date confirmed with no queue in front of it was never achievable, and the desk had been issuing dates the plan could not support. Move B costs nothing at all, because a refused short-notice drop frees a line-day that somebody else fills at once. Both moves are correct. Nadira ran Move A from January and added Move B in April, and the figure held above 95% for three quarters.

Then the book softened. By the third quarter of the following year, Velanthur had 862 booked line-days against the same 1,120 — 77.0%.

Neither rule noticed.

Move A, in an empty factory. The acceptance desk was still adding the queue allowance, because the rule said to, and the rule had been written when there was a queue. Velanthur quoted an average of 6.4 days later than it needed to. Three enquiries went elsewhere on lead time, 34,000 pieces each on average, at a contribution of USD 0.61 a piece. Contribution is what a piece earns after the costs that vary with it. USD 62,220 of contribution not earned.

Move B, in an empty factory. The desk refused 23 short-notice drops in the quarter, 3,120 pieces on average. In a full factory those line-days are worth more to somebody else. In a 77.0% factory they are worth nothing to anybody. 71,760 pieces refused, USD 43,773.60 of contribution.

USD 105,993.60 in one quarter, from a target that was being met. The monthly review saw 95.8% and moved on. Every month. Because the review was reviewing the target.

Why the target could be met at all

Look at what the two moves have in common. It is the general lesson, and it is not "people game targets".

Both moves make Velanthur do less, or promise less. That is not a coincidence. On-time delivery measured against a date you set yourself is not a measurement of delivery. It measures the gap between what you promise and what you do. There are two ways to close a gap, and promising less is much the cheaper one.

The proof is in the number nobody was reporting. Across the quarter Move A came in, the buyer's actual wait — from enquiry to goods on the water — went from 47.2 days to 51.9 days. Palliser waited 4.7 days longer for its garments while Velanthur's score for delivering them improved by 24.2 points.

The repair, which is one extra number

A target with one number in it can be met by moving either term. So Velanthur's target now has two numbers in it, and they pull against each other:

On-time delivery, Palliser's definition, at or above 95%, AND median quoted lead time at or below 44 days.

Move A now has a limit. The desk may add the queue until the quote passes 44 days. At that point the queue is the problem, and somebody has to decide about capacity rather than about dates. Move B stops being a delivery improvement at all and becomes what it always was: a capacity policy. Velanthur wrote the condition down beside it. Refuse short-notice acceptance only while booked line-days exceed 92% of capacity, tested on the first working day of each month.

That is the part you can carry anywhere. For every way your target can be met, name the thing that gets worse. If you cannot name one, the target is not measuring what you think it is, and it will be met by the cheapest route available. The companion number does not have to be in the same units. It does not need a target of its own. It mostly does not even need watching. It needs to be on the same page, so that a review which passes the first number has to look at the second.

Act as an operations director who has watched targets be met in ways nobody intended, and who is not interested in accusing anybody of gaming. I want to know every honest way my target can be met. My target is [STATE IT EXACTLY AS IT IS WRITTEN, WITH ITS THRESHOLD AND ITS DEADLINE]. It is measured as [THE FOUR CHOICES: UNIT, DATE BASIS, COMPLETION RULE, POPULATION]. The people measured on it can change the following things without anybody's approval: [LIST THEM - WHAT DATES THEY CAN SET, WHAT WORK THEY CAN REFUSE, WHAT THEY CAN RESEQUENCE]. My capacity position right now is [BOOKED UNITS OF CAPACITY] against [TOTAL CAPACITY] for the coming [PERIOD]. Do the following. First, list every distinct way the target could be met, including the ones that involve doing the work better. Mark each as either changing the work or changing the promise. Second, for each way, estimate the effect on the reported figure using my own numbers where I have given them, and say what you assumed. Third, for each way, name the thing that gets worse, and say who would notice it and where. If you cannot name anything that gets worse, say so plainly, because that is the finding. Fourth, for each way, state the state of the world in which it is the RIGHT thing to do, written as a number I can test monthly, and say whether that condition holds today at my current capacity position. Fifth, propose one companion measure that moves against the target, in whatever unit suits it, and say what it would have to be for the pair to be meetable only by improving the work. Sixth, write the target and its companion as two lines, with the condition underneath and the date it is next re-tested. Do not propose more than one companion measure, and do not recommend adding a target to any other department.