Predictive Analytics – why bother?

Guest writer: John McConnell, Applied Insights

In the on- and off-line worlds we are hearing more and more about predictive analytics (PA); analytical tools and methods which enable us to predict future events - for example to understand what stimuli different customers/visitors will respond to in the on-line context.

My own consulting company is one of those in the market trying to persuade digital property owners that they really should be applying predictive methods as a major part of the armoury which will move them into a more customer-centric mode.

But let’s be honest, if you are not already doing it, it probably requires a significant investment in different tools and skills … and it is likely that you will need to change many of your business processes to really maximise the benefits of that investment. So is it really worth the effort?

The answer is “it depends” but my firm belief is that many, many, more on-line businesses could gain marked benefits from PA. So why does it depend? … and how can we decide one way or the other?

First of all it is worth noting that, broadly speaking, PA comes in 2 flavours…

1. A packaged/productised form which can be more plug-and-play. These are the prix fixe dinners of the PA world
   
   
2. A more engaged/consultative style which is about applying expertise in a flexible way, often using more general purpose analytical tools, from the likes of SAS, SPSS and Oracle, to potentially address a wider variety of business issues. This is a more á la carte style.

It is usually easier to see if the prix fixe offerings have any potential in a specific business setting, and the on-line world has a growing number of examples of this class. Probably the most notable of these are TouchClarity and the multivariate testing (MVT) tools/services from the likes of Optimost, Offermatica and Memetrics. These vendors can usually tell us explicitly what level of improvement we can expect from using their tools based on their experiences with similar sites and sectors. We can then do the maths to see what that translates to for us in Return on Investment (ROI) terms, and the cost/benefit equation is quite straightforward. Their strength is that they have generalised quite specific applications of PA in a demonstrable and repeatable way (often combining technology and services).

TouchClarity and Optimost for example will be able to tell you what range of lift you can expect in your conversion rates and then will often proceed to prove that with a short, sharp engagement that minimises the initial cost but which enables you to build confidence within the business that “this stuff really works”. TouchClarity say that overall the expected conversion lift from their application is around 200%. Optimost talk of a “double digit” increase in conversion rates as the norm. In a recent article Seth Rosenblatt of Optimost gives the example of Time Life: “Time Life realized a greater than 70% lift in the number of products added to online shopping carts, which ultimately translated into a seven-figure impact on annual sales”.

When it comes to the á la cartes it is harder to see clear benefit for a specific business “off-the-page”. Hence it typically requires more effort for any organisation to assess the potential upsides. Having operated in this area for many years I would contest that the overall business benefits are usually the greater of the two styles and the investment of time and effort upfront, is almost always worth the expense. Where the prix fixes will give us quick, and often sustained, tactical wins (of significant value), the á la carte approach will allow us to take a more strategic view which often leads to a number of tactical projects with clear ROIs. Blue chip businesses like Harrahs, Tesco, Walmart and Royal Bank of Canada, who are demonstrable leaders in Customer Relationship Marketing (CRM), put PA approaches at the core of the customer understanding which underpins their success.

In our own consulting work we often take this approach to identify more strategic behavioural segments among on-line visitors (for clients like the Royal Mail and The Times). This kind of insight can inform a whole range of more tactical marketing activities, each of which will show yield improvement from the fundamental understanding of who the customers are.

Whichever of these two paths that you end up taking (and they need not be mutually exclusive) the central point is that buying into predictive analytics for the first time should be a 3 step approach which incrementally convinces a new adopter that it will bring tangible benefit. The steps are…

1. Find enough pertinent information elsewhere about success and failure to see similes within your own business. Do your think your business looks like Harrahs or Time Life for example?
   
   
2. Make an inexpensive internal review – based on 1) and perhaps with the help of consultants/vendors – to realistically assess the chances of success - or the risk of failure (“chance” and “risk” are exactly the metrics that PA gives us!).
   
   
3. Start to gradually prove PA to yourself and to the other stakeholders in the business. This usually means picking a good business objective in stage 2 - like optimise campaigns, improve landing pages to boost conversion, increase acquisition, etc. – and testing it.
   
   

So if you are not already doing so should you be doing predictive analytics? Well … it still depends … but it is certainly something that you should be seriously considering … before your competitors do.

John McConnell is the co-founder of Applied Insights, a UK-based consulting company specialising in customer analytics.

Return of the blog post soundtrack: Talking Heads - Once In A Lifetime

ShareThis