our views and our knowledge in analytics and other releveant topics

our blogs

New-born babies; A data-driven approach


I must confess that when the doctors presented me with what appeared to be a set of bagpipes full of milk, gas and excrement, to which someone had attached a medium sized turnip daubed with a crude painting of Winston Churchill, I was somewhat nonplussed. 6 weeks later, after much convincing and cajoling from my partner, friends and family, I am willing to accept that yes, he is a very tiny human, that he’s my son and that I’ll probably never sleep ever again.

Because of this turn of events, when the Station10 marketing team mentioned that it was my turn to write for the analytics blog, I was somewhat concerned that all my opinions on every subject had pretty much deteriorated to an “I agree with Daddy Pig” level. How am I supposed to come up with anything coherent and interesting for an audience wider than doting Grandmas, especially at 3am and covered in lacteous vomit?

Once an analyst always an analyst though, and as it turns out, certain aspects of customer analytics are very much like raising a new-born child. No really, bear with me on this – my son’s happiness and contentment is a goal, just like acquiring customers or making sales is a goal. There are a number of levers you can pull on to influence that sale (marketing channels, pricing, awareness, whatever), just as there are a number of things I can do to keep my son happy (which as a baby aren’t even that numerous – dry nappies, temperature, volume of milk, proximity of mummy and daddy, bounciness of surface).  So you pull on those levers and you watch the results come in, sale this time, no sale that time. Crying now, still crying now, crying more….oh god the crying….ahem. Sorry.

Anyway, the thing that most interested me was that I had started from a point where I figured that regular principles of analysing a relatively closed system (just in case you were worried, no, I don’t talk about my son like this on a regular basis) would apply to the baby.

Observe the data, find a pattern in his behaviour, address that pattern and bingo; a lovely, contented, mostly sleepy baby. Except there was no pattern, which was when my plan started to go off the rails.  To explain, let me take you on a brief diversion into the world of superstitious pigeons.

In 1947, the renowned behavioural psychologist B. F. Skinner conducted an experiment in which he raised pigeons in containers with no outside stimulus or influence, where, for a few randomly determined minutes each day, a machine would drop food for them. So far so animal protection baiting, but that aside, the results were quite interesting.  Three-quarters of the pigeons used in the experiment began to demonstrate ritualistic behaviours. Some hopped in specifically anti-clockwise circles. Some nodded their heads in a pendulum like fashion. Some switched back and forth between standing on one leg then the other, and so on. Skinner realised that the pigeons had conditioned themselves so that whatever action they were taking when the food had been delivered before, was what had caused the food to arrive. This drove the hungry birds to repeat it again and again, in hope of another small pile of seeds. Of course, the food delivery was completely random, nothing the pigeons did would have any effect on its arrival, yet they continued the rituals anyway. Skinner had made his pigeons superstitious.

This is what happened to me as a parent. Exasperated, but having addressed the primary concerns of the baby, having him fed, warm and dry, but still mewling, we started to fiddle in the margins. We put him on his back after he ate because that worked once. We gave him a dummy before we changed him because that worked once. We bought new mattresses. We only used the blue blanket. My dad, on one occasion, shaded his eyes from our living room lights one evening, at which point he stopped crying. On the next three times he sat with him, he held his hands awkwardly above the baby’s face for hours. It made precisely no difference.

So yes, that was a lot of words about pigeons and babies, and I’m sure you’re now going “get to the point already” so here are my lessons in customer analytics as learned from tiny babies and grey birds:

  • Systems are not as closed as they seem, you might think you know all the inputs and outputs of your business, but the fact is you don’t.  One baby human has enough randomness inherent to confound the best of us; your customer base might contain millions of horrible irrational fluctuating humans doing horrible irrational fluctuating things. No matter how much data you have, there’s more that you don’t, and a good analyst will always bear in mind where the gaps are.

  • Some parts of any system are going to be black boxed, because there’s no way your outputs can be complex enough to definitively prove causation by any given input – can you prove that television advertising definitively caused an upturn in sales? Even if you know that a customer watched an ad and bought your product there’s no perfect way to know direct cause and effect. Conversely, a good analyst should be able to put forward a good case, but if you take any conclusion at face value and don’t continuously test and learn, then you might well just be a superstitious pigeon.

  • Humans (and indeed pigeons) are inherently ingrained with the ability to see patterns, even (perhaps especially) when there aren’t any. The larger the data set being worked with the more patterns there are going to be. A good analyst isn’t one who can spot the patterns, it’s one who can spot the patterns that matter.

  • Take some time to step back. Businesses can feel like there’s a need to move as fast as you’d respond to the air-raid siren wail of a new-born, but gains are easier to identify the wider the picture. Sometimes you gotta go fast, but a good analyst will keep in mind that the quicker you go, the less rigorous the conclusions are going to be.

  • There’s a common perception that the gains to be made from analysing data on a larger scale are to be found in the nuances of that data, but make sure you’ve got the basics right first! Feed and change the pigeon before you start trying to wrap your customers in the blue blanket. Wait, I think I might have got something mixed up along the way there, but hopefully you see where I’m coming from.

So there you go, crying babies are like superstitious pigeons, and superstitious pigeons are like fickle customers, or something like that. If you want to know more about getting the most from data analysis, give one of Station10’s experts a call, just maybe not me right now, because I really need a lie down.