industrial decisions
Maybe time for the first bit of book promotion. “Decisions Nobody Made”, coming next year folks, is a book about … a lot of things, and one which it annoys me that I don’t have a simple summary like “Lying For Money is about fraud” or “The Brompton is about the folding bike company”. One of the things it’s about is the industrialisation of decision making.
But of course, that’s no good; “the industrialisation of decision making” isn’t a summary, it’s just a short phrase that needs a longer explanation. What does “industrialise” mean?
In context, industrialisation is something that happens to processes. If you’ve heard of “Taylorisation” or “Fordism”, it’s that, but it’s also what Adam Smith himself identified as “the division of labour”.
We should recall that the factory system significantly predates the mechanisation of production. History types have arguments about when one might date the Industrial Revolution, but the pin factory described in “Wealth of Nations” and immortalised on the old £20 note didn’t have any steam power in it; it was just a bunch of people in a room, passing part-made pins to one another.
Industrialising a process means breaking it down into simple elementary operations. But really importantly, in terms of the difference between an industrial and artisanal mode of production, the key thing about an industrial process is one of information and variety. Industrialisation works by standardisation. At any stage of production, there is no uncertainty (or at least, much less uncertainty) about what the input is going to be.
Each of Adam Smith’s workers knew that the next thing they picked up would be a part-made pin, of one of a small number of types, and that their job was to carry out one of a small number of operations on it then pass it on. When this is the case, you can optimise the ergonomics of the working space, build job-specific tools and frames, and make things hugely more efficient. (There’s a whole chapter in “The Brompton” about this, and a quote which sticks in my mind from their production manager – “to understand the jigs is to understand the factory”)
You can carry on this idea to think about industrialising a decision process … at a cost. Say you need to settle thousands of insurance claims every day, and you want to reduce the cost of doing so to a minimum. If you want to take advantage of the division of labour, you carry out the same process of standardisation and industrialisation. You design a form to capture the important data points, you draw up a flow diagram, define a list of checks and conditions for paying or turning claims down. If you don’t fancy this, a consultancy firm will be happy to do it for you and even write the software (or the SAP implementation). And the flowchart can be as clever as you like; you can even define flags that trigger an escalation of the process, to a specialised loss adjusting team that will gather more data.
As I said, at a cost. As Malcom Sparrow’s book “License to Steal” documents, this is how the US Medicare program got to a stage in the 1990s at which between a quarter and a third of all expenditure was fraudulent.
The problem is (as I said in “Lying for Money”) that when you standardise the process, you standardise the information set, and every decision about what information you’re going to collect is also, implicitly, a decision about what information you’re not going to collect. The industrialisation of the process creates a template, but it also creates a negative template for fraudsters. Sparrow called the method “shotgun, then rifle”, to describe the practice of sending in dozens of fraudulent claims of all kinds, finding out which ones got paid, and then sending in thousands of claims exactly resembling the few successes.
An obvious conclusion from this (which I would have said in so many words in “Lying for Money”, but I didn’t necessarily have the right language then because I hadn’t got so far into cybernetics) is that an adversarial decision-making process should never be industrialised. Sparrow actually prints a very good case study of one of the few Medicare insurers to have non-insane fraud rates, and they achieved it by not trying to cut costs in loss adjustment. They reasoned that paying the right claims was more important than reducing the average cost and of course they were right – a single fraudulent claim, at US healthcare prices, could outweigh a year of cost savings from process optimisation.
Less obviously, there are lots of non-adversarial decision processes where it’s also dangerous or destructive to standardise the inputs. At the very least, any industrialised decision making process needs to be monitored, to be sure that the world hasn’t changed too much. And it needs to be entered into in the understanding that unlike a manufacturing process, an industrialised decision process is here for a good time rather than a long time; the cost savings are temporary because the world will change and the process will need to be rebuilt. Understanding which decision processes can be industrialised anyway, and which ones are best kept open and artisanal might be the essence of management in the 21st century.