Optimality is in the Eye of the Beholder
I had a conversation recently with a colleague in which he remarked wryly about the many times he had met clients who expected to get ‘the optimal plan’, but had no idea what characteristics it should have.
His comment got me thinking — just what is optimality? From my experience, the answer is — it depends on who is answering it.
Ask an applied mathematician like me, and you are likely to touch off a long-winded Sheldon-ish response about how it all started on a warm summer evening in ancient Rome, and how the word ‘optimum’ has Latin roots, and how the Brits really were the first to start using it in a meaningful way during World War II, and… but you get the picture.
Ask someone at, say, a tire recycling company that uses/ should use route optimization, and the answer you get can depend on the role of that person in the organization. If it’s someone wearing a suit, they are liable to talk about minimized operational cost, minimized driving time & distance, maximizing the efficiency of vehicles used and so on. But if you ask a driver (who gets paid by the total weight of used tires he/she delivers to the recycling yard, at the end of each day)? Their goal is to pile on as many tires as they can, during each trip — and if the vehicle is already loaded to its theoretical capacity, but there are still 30-odd tires left at a stop where they have already arrived, are they likely to follow company guidelines and leave those 30 tires behind? Like hell.
Which might explain why in case after case, we would find that vehicles were routinely overloaded to 20% — 30% over capacity, despite company regulations (printed in VERY large font) that expressly forbade this.
Seriously, though, this is an important point in route (and) other optimization that is often overlooked. As consultants, there is a temptation for us to design the optimization system and its KPI’s to meet the requirements exactly as stated by the client, on paper — after all, they are the ones signing the check. But an optimization system that doesn’t get used isn’t worth anything, even if it’s bought and paid for. We need to talk to everyone involved in using the system, get to know their KPI’s and how they will be affected by changes the new optimized solution may bring into their lives, and adjust settings accordingly. All parties involved need to feel that the optimized solution is making their lives easier, and that they each have something to gain by using it.
In a past life, I used to work at an unnamed steel company (*cough* in Gary IN *cough*), developing optimization algorithms for inventory application and melt shop operations. I ran into the case of matching an order that needed coils of exactly the right chemistry as an unused slab in inventory, except that the slab was 20cm too wide. However, this particular order needed the coil produced by rolling the slab at the hot strip mill to be further treated downstream, at a pickling line — which had shears that could automatically trim coils to any desired width.
There were two clear solutions here — do it the way ‘it had always been done’, which was to
· from the slab yard, haul into the melt shop the pile of slabs that contained the desired slab
· unload slabs one by one from the pile till the desired slab was found
· load that slab onto a cutting table
· use an expensive acetylene torch to cut off the excess slab width
· load the shrunken slab back onto the slab pile
· take the slab pile back to slab yard
The alternative was to roll the overwide slab to an overwide coil, and trim it at the pickling line. (The trimming shears didn’t really care how much was trimmed off the coil width, as long as the coil emerged with a uniform width throughout.)
No-brainer, right? I mean, the amount of waste is about the same in both cases, and the company would save money and time by trimming at the pickling line.
Guess how long it took to get all departments to agree to accept the better solution of trimming at the pickling line.
Eighteen months.
Eighteen months of heated bickering and wrangling during several meetings, a couple of which nearly came to fisticuffs.
The reason? One of the KPI’s that the manager of the pickling line was measured on was the amount of waste steel that his department produced — and trimming that overwide coil at the pickling line would ding this KPI of his. The compromise solution that everyone grudgingly agreed to was to have the melt shop (which would have been penalized for the waste produced by cutting the slab to size) compensate the picking line by agreeing to take the hit for the waste steel produced by trimming the overwide coil.
In conclusion — if and when you come across a document that speaks of an ‘optimal plan’, examine it with a jaundiced eye. And make really, truly sure exactly what that ‘optimal plan’ is expected to achieve.
References
1. ‘It is a warm summer evening in ancient Greece…’ https://www.youtube.com/watch?v=2XZQ3gympgQ
2. Origin of ‘optimum’ https://www.merriam-webster.com/dictionary/optimum
3. British usage of operations research in WW2 https://pubsonline.informs.org/doi/pdf/10.1287/opre.35.3.453
