Interactive Statistics and the Scalability of Web Analytics Practices

I apologize in advance for the lengthy post: I’ve been in business plan purgatory for what feels like 6 months.

One of the greatest challenges to any web analytics practice over the short and medium runs is the base skillset of interactive statistics.

First, a definition of interactive statistics.

Interactive statistics is the body of knowledge that is based around the derivation of insights out of web analytics and/or business intelligence software.

A brief rant.

I only really got into statistics in high school math honors. The New Brunswick honors curriculum emphasize practical applications of statistics (as opposed to traditional LaPlace style urns and cards) to the real world. Those who really know me well frequently chortle that I’m capable of outstanding memories around confidence intervals, yet sometimes find it hard to mentally tell the difference between 9×3 and 4×7. (Yes, yes, and don’t start saying an odd times an odd is always an odd. I could hardly prove my way out of a paper bag in discrete math).

One of the core questions that really opened my eyes revolved around how precise I needed to be, as a product quality control engineer, if the government mandated the number of pills in a bottle had to 95% of the time at a standard deviation of 3 grams.

Later, I’d get into statistics the political science route – always emphasizing the practical notions of statistics, while making sure that I’m learning the hardcore theory. (And, I thank my of my cohorts who were doing advanced statistics in the mathematical science departments.) But, the spark didn’t originally come from university – where most Canadians are introduced to real statistics for the first time…it came from high school.

And I don’t think that most high schools emphasize applied commercial mathematics.

So, that’s the rant.

I don’t think that North American is producing many functional statisticians.

By that, I mean, people who have taken 2 or more applied statistics courses in their final year of university, and then having a chance to apply them.

Next, let’s layer on the “Interactive” part.

There’s an entire world of information architecture, pathing, and a set of practical and theoretical literature around how people interact with websites. There is also an entire set of definitions – for instance, “time spent on site” doesn’t actually really mean “time spent on site”. So, there’s a definitive “interactive” aspect.

Layer the both together, and you what we consider to be a solid “A” analyst – Statistics + knowledge of interactive.

This is one of the core skillsets of a specific flavor of web analysts.

Not all web analysts need to be statisticians. There are web analysts that are purely strategic varieties. There are also solid communicators that are important. There are also technical web analysts, and a new flavor that is a ‘qualitative’ web analyst.

However, almost every team should have a few interactive statisticians…these are the people who can extract a different type of insight from the data. (There are non-statistical methods to be sure).

However, finding these people is hard. It’s even harder to find people with a passable skillset plus the required curiosity and tenacity it takes to be effective in these early days.

A lack of relevant skillsets in the labor market is the greatest problem in the scalability of web analytics practices.