Understanding the role of behaviour and predictive analytics

Companies are focusing on drawing real-time online audience insights, and enhancing their understanding of how consumers perceive and engage with their brands, content and marketing.

And they are delving into various techniques which allow companies to understand and predict customers’ desires and to more effectively serve relevant content and products in real time. 

In order to know more about these techniques, EyeforTravel’s Ritesh Gupta recently spoke to Expedia’s director of business development EMEA, Cameron Jones  and Vicky Brock, co-founder, Highland Business Research. Excerpts: 

Can you provide an insight into behaviour analytics? How is it measured and how does it help in optimising website content?

Cameron Jones: Behaviour analytics help us understand and predict our customers’ desires and to more effectively serve relevant content and products in real time, ultimately increasing satisfaction and conversion.

For example if a customer is searching for a hotel in a gateway city for a two-night midweek stay, this is likely to be a business customer that favours certain types of hotels and is focused on making a quick purchase decision without the disrupting of ads…behavioural analytics helps us to initially segment and understand certain customer groups and then anticipate new customers needs when they display similar behaviour on the site. The sort order for hotels is a critical element of providing relevant hotels to customers. The hotel sort order at Expedia has benefited from years of development and utilises sophisticated algorithms which include behavioural data to display the most relevant hotels to customers.

Vicky Brock:  Any decent analyst is looking to identify how different segments of users behave on a website.  It is important to remember that not everyone coming to a travel agent site, for example, has come to buy.  Some are at the early research stages and are trying to figure out where to go.  Some are at the stage where they know where they want to go and are making price comparisons. And others have already booked and looking for specific information.

All these groups exhibit very different behaviours on a site and all have different success outcomes associated with them.  Some may not even be worth considering in the data.

Understanding why people come, how they behave and how they achieve purpose of visit requires an integrated suite of tools. But ultimately this is what users are trying achieve with an analytics programme:  Improve business performance by helping more users achieve what they came to the site for.

The next step is behavioural targeting.  When you’re confident enough about your data, segments and can recognise behaviours associated with those segments and specific purposes of visit, you can automatically begin to serve specific, relevant content to those people.  

If you can recognise I’m returning to a product that I bought 6 weeks ago, isn’t it better to try and sell me insurance, trips and other add-ons, rather than to show me that the product has now been discounted further?  Good segmentation, combined with multivariate testing and friendly web developers means you can at least make some inroads into customising what people see, based on how they behave.

This is one of the areas where in the coming years I thing the web will really change.  I think in the comparatively near future the one size fits all website will be dead.

What do you make of the usage of predictive analytics to mine and analyse the data gathered from web traffic and bookings on a website? 

Vicky Brock:  Once you’ve got really good at looking out of the rear window, your attention has to move to looking forwards.  This can be hugely complex and it’s an emerging analytics field in its own right, but it can also be done with existing analytics and booking data.

I’ll give you an example.  For a ski resort that has a high trafficked website, but most of its tickets sold offline, we have been able to use page view data to specific parts of the site and look at the correlations between offline sales.  By doing some cool maths, we have been able to determine that if conditions are fine, it is possible to predict how many people will show up Saturday based on the traffic on key pages of the website on Monday/Tuesday.  This means not only is it possible to better predict staff requirements, there is also a 3 day marketing “save” period if numbers look like they’ll be down.

Cameron Jones: Predictive analytics help to prioritise projects and investment in resource with expected revenue and profitability impacts. Each project we embark on has clearly set out objectives, goals and expectations which we measure against pre and post shipping an enhancement. 

Predictive analytics helps us understand the impact of enhancements on various parts of the business and enables us to plan more effectively search engine marketing strategies through to call centre staffing levels.