Outliers - classic BI

In a previous Industry Insight, I discussed how the book “Freakonomics” (by Levitt and Dubner) provided insights into how data can translate into entertaining and influential stories about human behaviour. The point of this was to show how this is a great approach to look into, for anyone considering or already in a career in the business intelligence (BI) field.

Similar lessons in BI can be found in Malcolm Gladwell's 2008 book: “Outliers: The Story of Success”. Both books mentioned were bestsellers to a broad audience who would otherwise not necessarily buy a book about economics or statistics. As an aside, they provide lessons in how the way BI results are presented can be just as important as the results themselves. In many instances, the way numbers, results or data are presented can change the way people perceive them.

In “Outliers”, one of the first insightful stories is an investigation into why an unusually high number of elite Canadian ice hockey professionals were born in the first few months of the calendar year. The presumption would be their birthdays should follow the more “normal” distribution pattern throughout the calendar year, from January to December.

In the book, the author's conclusion is that because youth ice hockey leagues grade by age, then in every year, players born on 1 January are playing with players born on 31 December. For younger players, the one-year age gap can make a significant difference in size, speed and ability, which makes the likelihood of being selected for a team better for the relatively 'older' players in this regard.

Outliers are the BI equivalent of the “canary in the coal mine

David Logan is principal consultant at PBT.

Being selected means more game time and more coaching, which ultimately leads to further improvement, all culminating finally in the 'outlier' that disproportionately more pro hockey players are born in the first few months of the calendar year.

Ironically, this is a classic BI story - the data leads to outliers, the outliers initially have no immediately obvious causal effect; however, by applying more than just number-crunching, ie, behavioural analysis, observation and intuition, then really interesting stories begins to develop. On a more practical level, an example would be holding back children at school a year in South Africa when they have summer birthdays (so they could be slightly younger for their grade or older for their grade), which could be beneficial to their overall performance.

The key learning to take away from reading this Industry Insight is that outliers are critical techniques in BI. Reporting the top-selling salesman/product/sales channel is the “bread and butter” of BI. Identifying the outliers, such as the relative quality of sales (in terms of repeat sales, customer lifetime value of sales, disproportionate cost of sales) is moving more towards BI than reporting.

Building a story of why these outliers exist (a certain salesman may be selling “at all costs”) is real BI, which challenges and changes mindsets in an organisation to add value to the bottom line. In most cases at clients, outliers are indicators of previously unrecognised over- and underperformance. In addition, the absence of an immediately obvious reason for these outliers usually means there is one, it just hasn't been found (yet).

In fact, this is one of the most valuable things using outliers can contribute to a business. The mere existence of an outlier focuses attention, often, on the most valuable area of a business, without (initially at least) actually knowing quite what it is that is going to be found. It could be said that outliers are the BI equivalent of the “canary in the coal mine”. Miners used to carry canaries in a cage, and when the canary dropped dead, they didn't know which gas or how much killed it, but they knew it was time to leave the mine.

In essence, outliers have the potential to enable businesses to identify their most valuable aspects, and thus can offer businesses the opportunity to improve on the aspects where they are still 'lagging'. What this ultimately boils down to is the fundamentals of successful BI.