Human vs Algorithm - Test Yourself!

‘When people think, they simulate the world.’ This is a quote from Jordan Peterson’s ’12 Rules for Life’.

When we have enough confidence in our perception (or ‘simulation’) of the world and the way it works, we act upon it - and action is key to business. However, the foundations for our confidence may not always have been fully explored or well formulated and some people are guided mainly by instinct and ‘common sense’ (not always so common).

But the demand for us to look at reality in depth – to get ‘beneath the hood’ - has grown. The demand to apply logic, calculation and computing power to this analysis has grown. In short, the demand to model our world has grown.

Because models augment the human capacity for thinking.

In business, Excel models are ubiquitous. Using models helps us to make better decisions faster, and to dismantle the reasoning behind our decisions step by step all the way back to initial assumptions and data. Models can handle a great deal of complexity very quickly and they can derive answers to routine problems that would otherwise just baffle us.

Not convinced? Well then - try this little online game I built: https://bit.ly/2LQMo6Q.

 This game’s premise is quite simple:

 1.) You are a reseller who has some cash in the bank and a certain number of units already ordered to meet sales already secured for the next 15 months. Goods cost £5/unit and sell for £8/unit, sales take place 2 months after goods are ordered, and the profiles by which you pay creditors or receive payment from debtors are known.

2.) Suddenly, you learn that you will be able to sell everything you can order: all you have to do is buy as many goods as you can over the next 15 months whilst make sure that your cash balance does not drop below a predefined limit.

3.) And, for model sceptics, competing with you will be an algorithm doing the same thing - maximising purchases whilst staying above the minimum cash balance.

Should be easy, right? Perhaps not all that easy to maximise the performance as the algorithm does, but at least to get within 80% or 90% of its answer, surely? After all, we are aware of the dangers of overtrading and how not managing working capital properly can be disastrous, so a little good judgement should go a long way…

Well, try it and see.

The game is constructed to give you good feedback (such as in the bold blue text in the below picture) when you get within even 50% of the algorithm’s answer. That’s right: you will get praised for being just half as good as the algorithm.

(As a tip: if you are playing this on your phone, tilt it horizontal so that the graph displays properly, and click the ‘Show Balance’ button to view the forecast cash balance before selecting how many additional goods to buy.)

Much of my experience has involved consulting in financial problem solving – and building models (predominantly Excel, and deterministic - not stochastic).

Such consulting engagements often look like this:

1.) There’s some defining of what the question or task actually is, followed by planning of what needs to be explored and analysed, how it should be analysed, who should be consulted, and when tasks should be performed.

2.) Then there’s the building of a model – the equivalent of writing a first draft report, or painting the outline of a picture: something that can be shared, but within a small group of constructive critics, not fit for public distribution.

3.) Finally, there is the finished product - a user-friendly, flexible tool which take in a mass of information, feeds it through a web of logic and calculations and then outputs an array of graphs, tables, financial statements, warning flags and other indicators.

The final model is a tool that enormously enhances how a business story is told, adding robustness, depth and simplicity.

So, if my client is under pressure, a model can act to carry some of the load. It can generate information that is demanded, and it can be the first port of call for sternest critics bent on exhaustively interrogating the reasoning and data that underlies a financial result.

And smart managers know this. They want to use anything and everything that will augment their business performance. They want the smarter, faster, more powerful model. With bells on.

Model design as a field is developing – in fits and starts at times, and in leaps and bounds at others.

Maybe in the next few years, models will even learn to ask questions and think for themselves. It is already possible to embed some algorithmic machine learning / AI resources in models and workflows (look at IBM’s Watson offering).

A great amount of human focus is directed toward financial problem solving, to developing the personal skills and abilities to do it well, and to mastering the tools which enable it.

Models are such a tool, so get the best models you can and use them well!