The Anna Karenina principle describes an endeavour in which a deficiency in any one of a number of factors dooms it to failure. Consequently, a successful endeavour (subject to this principle) is one where every possible deficiency has been avoided. The name of the principle derives from Leo Tolstoy's book Anna Karenina, which begins “Happy families are all alike; every unhappy family is unhappy in its own way.”
In this blog post I want to focus on one particular phase of the predictive analytics process – deployment. As the Anna Karenina principle suggests, there are an infinite number of ways in which predictive analytics projects can fail, but it’s fair to say that deployment is one of the ones that crops up most often.
I’ve been involved in a wide range of discussions recently with both clients and prospective clients about adoption of advanced and predictive analytics in their organisations. Despite wide variety in the nature of the individual organisations’ activities, their purpose, their priorities and their specific challenges there is always a point at which results must be deployed. Without successful deployment all the time, effort and money that went into the analytics phase is wasted, so it’s really critical to get the deployment phase right.
By deployment we really mean the ability to put the fruits of your analytical labours to use in the day-to-day operations of your business. If you don’t plan for and consider this deployment phase carefully, the chances of reaping significant and sustained benefits from predictive analytics diminish very rapidly. It’s particularly important to have some idea of what deployment will look like for your organisation before you start work. This can take many forms.
Perhaps your marketing team would like a list of scored best prospects, selected and ranked by those most likely to respond to a planned campaign, in their hands before they send out a campaign. Maybe you’re a utility company and you’d benefit from a scored database of fixed assets that specifies the likelihood of each failing within the next 30 days. You could have a website and want to deliver content intelligently, to an unknown website visitor, which will match their current interests based on the last 5 most recent clicks they have made on your site. Alternatively you might want to score callers to your call centre in real time, based on the likelihood that each one will terminate their contract with you during a given time period. Perhaps you want to go further than this and use predictive analytics to give you two recommended intervention strategies that are most likely to be effective, in reducing this churn risk.
The point of these examples is not how sophisticated or complex they are or not. The point is that they each represent real and credible examples of predictive analytics being deployed in a business. It is all too easy to set out with wonderful intentions of how predictive analytics could help you improve your business. The reality is if you don’t think the practical applications all the way through to deployment your predictive analytics project is likely to be another fulfilment of the Anna Karenina principle.