Why personalisation is like detective work

Why personalisation is like detective work

Personalisation on the web is rather like detective work.

However, just like in all good cop shows where the innocent guy gets taken into custody, not many brands gets personalisation right first time. Even Amazon (who invest millions on data modelling) are still chasing me around the internet with ads for condenser microphones and Lego. I may buy more bricks, but I won’t be buying another mic – not all products are equal.

To avoid a miscarriage of justice (and to sell more products), you first have to build up a decent case. In this instance, it means using data and inference. This involves:

  • drawing up lists of likely suspects (e.g. people who will be more likely to be using your website – this will often come for customer research of marketing personas);
  • using forensics to eliminate lines of enquiry (e.g. using data from analytics); and
  • interpreting suspect motivation and predicting behaviour (e.g. information from surveys, motivational psychology).

There seems to be a clash of expectations, though, between brands aspiring to serve their customers with personalised experiences, and customer attitudes to the way that it is executed. Adobe (a respected provider of technology to enable personalisation) found that only 33% of users thought that personalisation was valuable. CIM's 'Whose data is it anyway?' report found that 70% of consumers don't see the benefit of sharing their personal data. A further study by Lyrus found that 37% of users appreciated being targeted, but 31% did not appreciate. And finally a study by Janrain discovered that 77% of customers would trust businesses more if they explained how they were using personal information.

It is also clear that there is some suspicion around data-driven marketing…transparency is key. So, when you start thinking about personalisation for your own site or your clients’ sites, you need to consider ‘what is appropriate familiarity?’  

There are plenty of examples in the offline world where a little personal touch goes a long way. Let’s think about a few of these:

  • You go into your local to meet a friend and they’ve already got your pint in.
  • You go into a store and the shop assistant makes recommendations – sometimes they don’t suggest the most expensive product.
  • You buy a present in a department store and the shop assistant gift wraps it for you.

Little touches like these make a big difference. And it should be the same in our digital lives.

All of these examples are timely and appropriate. Where brands fail is by trying personalisation in all circumstances and at all times, conflating personalisation with interruptive marketing or using incomplete data. An example of inappropriate personalisation going disastrously wrong is when Target (a US retailer) sent coupons to a female based on her purchase behaviour with discounts for pregnancy related items. The story broke in the New York Times that the retailer had predicted her pregnancy before she knew herself.

The other risk of personalisation is not exposing customers to new products or services outside of items you know they are interested in. Filter bubbles such as these narrowly reinforce a defined list of choices. So, although it looks like the personalisation is a success, you may be at risk of not exposing users to a greater variety of content leading to a richer experience.

1. Get a good toolkit.

There are many pieces of technology that can get you started that will fit into your available budget. At the enterprise level, tools like Adobe, Optimizely and Monetate offer powerful personalisation engines via an easy to use interface. But there are less expensive offerings too, including Bunting, AB Tasty and Omniconvert, which have pre-setup personalisation segments for factors like geo-location, weather and time of day.

2. Define your objectives and key segments.

Find out whether a personalised service fits specifically with your business. What is it appropriate to personalise? You need to gather data from multiple sources, including web analytics, Add to Wish List features, social shares and favourites, and reviews on your site and others.

3. Where should personalisation appear on your site?

Use analytics to tell you where users are abandoning the site, and to calculate what are your most valuable pages or funnels.

Querying your site search data will give you an indication of the most searched for products or product types, while Geo IP data will provide you with your most important geographical locations.

4. Leverage user information.

Chances are that there is a lot of valuable knowledge locked up in your business, so make sure you are regularly having conversations with people in your customer services team who are fielding client queries day in and day out. Supplement these with exit interviews using tools like Qualaroo, and regularly user test new features on online testing platfroms like UserTesting.

5. Finally, use live online split testing with real visitors to try a few things out.

What is an appropriate level of familiarity? Undertake A/B testing on offer straplines or email copy, as well as pricing, cross sells and upsells. And after you’ve done this, analyse and repeat.

In summary, personalisation presents a great opportunity to build customer relationships. However, you need to understand what is and is not appropriate. Complete personalisation is impossible – and isn’t always desirable. Your customers will give you data on how, where, when and why they are happy to be sold to, as long as in return your business is honest and transparent.

To hear more from Joe Doveton, sign up to our Conversion Rate Optimisation in a Day course. The introductory session will explore how introducing testing and measurement can help you transform your web business. 

Joe Doveton Course Director CIM
Back to all