how and why to lie with spreadsheets

a form of communication that can't be used to lie with isn't worth doing

A while ago, while doing research for “Lying For Money”, I read a book about pirated decryption cards for pay-TV.  I ended up not including this as one of the stories, because it’s quite technical, and it involved quite specific allegations against people with scary lawyers which I wasn’t in a position to confirm, but one of the themes of the book has stuck with me. 

That was from its description of the early days of smartcard hacking.  One of the people involved in that scene reminisced that because the pay-TV providers were always trying to cut costs, they tended to go through a cycle of installing a protection system, waiting until it got compromised and then rolling the next most expensive solution.  He suggested that effectively, by always making small incremental improvements to security, they were effectively running a training course for hackers.

As I say, it’s stuck with me because it feels like a concept of general application.  Something similar is at work in the famous Medicare fraud strategy of “shotgun, then rifle”.  If you provide feedback on unsuccessful attacks, then an adversarial party can use that feedback to quickly learn what works.

But the concept is probably even more applicable to non-adversarial, or perhaps semi-adversarial contexts.  If you have a set process for justifying decisions in a company or a government department, then it is quite difficult to organise that process in such a way that it does not become a training course for your employees in how to lie to you.

Everyone in business knows this; everyone in the investment world knows it even better.  If you are half good at your job, you can fudge a spreadsheet to beat a hurdle rate or a payback period.  Any merger or acquisition can be earnings accretive if you try hard enough.  If you are properly good at your job, you can hide the key assumptions so cleverly that they’ll never be found.  (I occasionally say that in discounted cash flow models, dilettantes fudge the cost of capital, amateurs fudge the growth assumptions but professionals fudge the tax rate).

It could even be said that if you don’t know how to get the result you want (or the result your boss wants) out of a spreadsheet, you don’t understand the business well enough.

Which sounds like one of those half-clever pieces of fortune cookie cynicism for which I used to be renowned.  But I think it’s actually true, and it conceals some important points about how decisions are made which are worth thinking about more closely.

Every accounting system is a mental prison, to take the title of the first ever post.  The raw material of any spreadsheet is itself a lossy representation of reality, put through an information-reducing filter.  You then take a subset of that lossy representation, and make it into a model, further attenuating the information.  And then you bring the model along to the investment committee, or similar.

As a manager, at this stage, your very job is to bridge the gap between this reduced representation and the underlying reality; to put that information back in.  Futzing around with the model until it gives you an answer that makes sense isn’t “gaming the system”, it’s doing what you’re paid to do.  If you don’t understand how the inputs of the model relate to the outputs and you can’t reconcile a set of stylised adding-up constraints and simple extrapolations with your understanding of the right answer, then yep, you don’t understand the business well enough.

Spreadsheet models (and many other kinds of models) are a kind of rhetorical device.  They’re a way of making a particular kind of argument.  Specifically, they’re a way of quickly showing other people that your argument can be made consistent with the accounting system of your decision making unit, without having to show your workings cell by cell.  But they’re a means of expression; they aren’t, can’t be and shouldn’t be the argument itself.