The Stat Man

Author: Geoff Walsh

The joys of analysing and collecting statistics!

I love statistics. There is something unerringly pure about unadulterated numbers. I joined Intelligent Environments as a support technician, but soon found myself collecting and analysing system statistics, mainly for the fun of it. And before you laugh, the point of this blog is to let you in on a fascinating secret: for the vast majority of the systems we host, customer growth (i.e. the number of people who use the systems) is almost exactly linear! Take a look at this graph of one of our datacentres:

graph1jpeg_500x255

Examining our datacentres

This shows the total number of users who actively log into the various systems hosted in one of Intelligent Environment’s datacentres over a period of just under a year. This includes systems for card servicing, bank account checking, collections account checking, mobile apps, mortgage checking, etc. There is a dotted ‘line of best fit’ in there somewhere, but so close is it to the actual line that it is barely visible. Indeed, for those statisticians amongst you, the R2 value is 0.9956. (Basically the nearer the value is to 1, the more linear the data, where 1 means exactly linear; to put it in context the R2 value for global warming for the last 15 years is 0.04.) .

See, told you it was interesting.

What this means is that despite the various systems we host, aimed at different people with different needs, different systems for different reasons, human beings are so predictable that we can say with 99.5% certainty how many of them are going to be using the system on a certain date in the future.

However… what we cannot say with such certainty is how often these users will log on. Here is a graph of the maximum logons per hour per day over the last year for a single customer:

grpah2jpeg_497x290

Studying customer behaviour

The low points are Sundays, but the high points? Seemingly random, but a closer look shows that barring bank holidays, each week follows a pattern – a spike on Monday and Friday and a lull in between. This makes it a little bit easier to predict, until one day in November BOOM – we saw a large spike in users deciding to log on at the same time. No doubt there are reasons for this, but from our point of view it was out of the blue. This shows us that predictions of customer behaviour can only be taken so far, and that systems still have the capacity to surprise us. Hence these systems have to be designed to cope with far more than the maximum predicted usage.

Clever banking predictions

There is a whole debate to be had on the reasons driving these two disparate but linked statistics (Psychology? Instinct? Clever marketing?) but for now I’ll leave you with the thought that I know when you are next going to check your bank balance…

03 Dec 2014

Author: Geoff Walsh

The joys of analysing and collecting statistics!

I love statistics. There is something unerringly pure about unadulterated numbers. I joined Intelligent Environments as a support technician, but soon found myself collecting and analysing system statistics, mainly for the fun of it. And before you laugh, the point of this blog is to let you in on a fascinating secret: for the vast majority of the systems we host, customer growth (i.e. the number of people who use the systems) is almost exactly linear! Take a look at this graph of one of our datacentres:

graph1jpeg_500x255

Examining our datacentres

This shows the total number of users who actively log into the various systems hosted in one of Intelligent Environment’s datacentres over a period of just under a year. This includes systems for card servicing, bank account checking, collections account checking, mobile apps, mortgage checking, etc. There is a dotted ‘line of best fit’ in there somewhere, but so close is it to the actual line that it is barely visible. Indeed, for those statisticians amongst you, the R2 value is 0.9956. (Basically the nearer the value is to 1, the more linear the data, where 1 means exactly linear; to put it in context the R2 value for global warming for the last 15 years is 0.04.) .

See, told you it was interesting.

What this means is that despite the various systems we host, aimed at different people with different needs, different systems for different reasons, human beings are so predictable that we can say with 99.5% certainty how many of them are going to be using the system on a certain date in the future.

However… what we cannot say with such certainty is how often these users will log on. Here is a graph of the maximum logons per hour per day over the last year for a single customer:

grpah2jpeg_497x290

Studying customer behaviour

The low points are Sundays, but the high points? Seemingly random, but a closer look shows that barring bank holidays, each week follows a pattern – a spike on Monday and Friday and a lull in between. This makes it a little bit easier to predict, until one day in November BOOM – we saw a large spike in users deciding to log on at the same time. No doubt there are reasons for this, but from our point of view it was out of the blue. This shows us that predictions of customer behaviour can only be taken so far, and that systems still have the capacity to surprise us. Hence these systems have to be designed to cope with far more than the maximum predicted usage.

Clever banking predictions

There is a whole debate to be had on the reasons driving these two disparate but linked statistics (Psychology? Instinct? Clever marketing?) but for now I’ll leave you with the thought that I know when you are next going to check your bank balance…