Tailor Made

Newspapers pride themselves on knowing their readers and invest substantial sums in ensuring that continues to be the case. Fail to deliver what the readers want and circulation swiftly drops. With the advent of the internet, many hoped that the process of providing relevant editorial could be taken one step further with the introduction of personalised content delivery. Readers’ preferences would be catered for at the most minute scale with tailored editorial, advertising and promotions served up on every site visit. This would be a mutually beneficial arrangement as readers would receive editorial content that interested them and the site publisher could accurately target advertising and promotions. Genuine one to one communication on a mass scale had finally arrived. Or had it?

Unfortunately, as with so many internet ‘solutions’, the technology lagged some way behind the good intentions. Early offers of ‘my home page’ usually provided the users with the options of selecting from a preset range of content streams. This is personalisation of a sort, but nowhere near the sort of tailored product that the early innovators had in mind.

If the users play along with site operators and fill out user registration forms, it is obviously easier to gain insight in to their likely preferences. Few people, however, will fill in a registration form unless given a good reason to do so. Deriving useful data from less cooperative users, based purely on the way they interact with the site has always been the challenge. In theory, it should be possible to make meaningful inferences about these users by analysing their online behaviour and then act on this information - in real-time - to provide tailored content. Until now, this has proved to be too onerous a task for all but the most determined of site operators to undertake. But this may now be changing.

Finally, the technology - in terms of content management systems, real-time behavioural analysis and data processing - is beginning to catch up with some of the early promise. Sites are edging closer to the goal of offering a tailored online experience to their users, with all the commercial and editorial benefits this more personalised browsing would allow. However the degree to which this is possible, or even desirable, varies a good deal depending on business and content model employed.

Ad Serving

The technique of serving up appropriate banner advertisements, based on broad analysis user behaviour, is already well established, allowing website owners to replicate the sort of placement that can occur in the print model. Equally well-established in the ecommerce sector is the practice of designing site architecture that can be manipulated to ensure that the user is moved efficiently towards the goal of achieving a sale. While websites in the media sector have taken swiftly to the banner ad model, few have embraced the possibility of tailoring the online experience through manipulation of content. But this should now be possible given a robust online data analysis tool combined with an equally robust content management system.

Tracking User Behaviour

"You need to determine where a user has come from, what content they have looked at and what people who looked at similar content did next," says Steve Wind-Mozley, managing consultant at web analytics firm Nedstat. Basic analysis derived from recording the way the user interacts with content means that the user can be placed into a market segment and predictions made about what they may wish to do next. "Analytics allows you to ask the question, ‘what do I know about people who consume that content in that order?’" says Wind-Mozley. "You can then set up business rules which trigger the content management system to serve particular content when the user behaves in a particular way."

Analysing a user’s route through the site to determine where they lose interest is also a possibility. This allows the site operator to tweak the site architecture, so that the user’s interest is maintained. "We can carry out what is known as a champion challenger testing to see which pages work best and in which order to retain the user," says Wind-Mozley.

The Funnel

Website builders - particularly in the context of ecommerce - will often look at their creations in term of a funnel. The intention is to funnel users towards a desired action, ensuring as few as possible will be lost in the process: "Analysis allows you to optimise the funnel to stop leaking, so for example if the site involves filling in an insurance form and people drop out when they are asked to estimate the value of their glassware, you would remove that question and stop that leak," says Wind-Mozley.

Certainly the comparison of a website with a funnel is an apposite one. The home page is easily the most visited part of any site, but traffic rapidly trails away further into sites. For any publisher, encouraging footfall in less popular parts of a site and utilising all available content has to be a priority. "With any portal, as you get deeper the audience gets very thin very quickly," says Donald Hamilton, managing director of Wunderloop UK.

Wunderloop’s behavioural targeting technology uses data gleaned over repeated visits to predict what content a user may be interested in, gradually depreciating old viewing habits to ensure the content remains pertinent: "We work out what should be served by weight of relevancy," says Hamilton. "What a user is viewing right now is 100 percent relevant, what they looked at on their last visit 80 per cent relevant and so on."

Repeated visits obviously allow a more accurate picture of a user’s behaviour to be built up, but Hamilton maintains that within a relatively short time, valid assumptions can be made: "We can learn something within three clicks by a user. On the second click we have started to understand what they are looking at and at the third click we have a good idea of what they are interested in, in that user session."

Tracking Methodologies

Wunderloop tracks users via cookies; other systems will make a record of a user’s IP address. Critics suggest that reliance on these techniques can hamper any such attempts to track users and provide a tailored online experience: "If a user comes from a corporation then they will have a blanket IP address applying to many users while many ISPs now generate a new IP address every time they log on," says Bill Murray, group business information strategy managing director at Haymarket who is sceptical of the benefits of serving tailored content.

Hamilton accepts that placing cookies and recording IP addresses can prove a stumbling block and concedes that for this level of analysis to succeed, a closer bond must exist between user and site. "We work with partners where there is a high degree of user registration and people trust the site," he says.

Peer Review

Exploiting a user’s interaction with a site to build a product tailored to their interests doesn’t have to rely on stealthy observations. At Future Publishing, the company has built a bond of trust between user and publisher that has led to some very practical editorial results. The company publishes Windows Vista magazine and its online incarnation utilises technology to exploit the growing demand for user generated content and provide a degree of personalisation. Once a user has registered, he or she can submit articles which then have the same chances of appearing on the site as pieces produced by the magazine’s journalists. Articles are voted on and the number of votes received determines how prominent a position that article will occupy on the site. The software also records the type of articles a user votes for and serves up similar items to their home page. "We wanted to build a community around the operating system and this system allows users to share tips," says editor Paul Douglas. "The consumer becomes the filter so only the good stuff rises to the top."

The website undoubtedly operates on a very egalitarian model but is it valid in a wider context? Whilst acknowledging the democratic virtues of Windowsvistamagazine.co.uk, Bill Murray remains to be convinced that there are advantages in serving up editorial content based on observed behaviours or preferences: "There is a danger that it can become too automated which is to assume that the spontaneous nature of the web has no merits," he says.

Murray argues that the traditional methods of sifting and serving content to users should not be discounted: "We (Haymarket) offer content modules users can select to appear on the home page. Many people interested in specialist areas find this a useful tool, but there are plenty of others who are happy to rely on good old fashioned methods and let editors make the choices," he says.

Whether or not Murray’s sentiments remain centre stage may depend on the competency of emerging technologies and the return on investment they provide. Already there are indications of where the market for online analytics might go next.

And, either, or

Relying on keywords to indicate where a user’s interests lie can still be a tricky business as this does not take account of the context in which that word appears - hence the odd results search engines sometimes deliver. Soon there could be more intelligent eyes focused on browsers. Technology being developed by Swedish firm Hapax dispenses with the methods of traditional Boolean search engines for a more sophisticated analysis: "We use natural language search technology to look at the root of a word, the part of speech it appears in and any modifiers or adjectives acting upon it," says Mark Redgrave, CEO of Hapax UK.

Employing more sophisticated language analysis can, for example, avoid the risk of an advert for an airline appearing on a page carrying news about an air crash. More importantly, it can help to identify preferences and enhance any segmentation process: "If a user arrives at a site based on the search phrase ‘I hate skiing in France’, this does not mean that person hates skiing or hates France. Understanding this preference is fundamental to accurately profiling that user and a lot of this rich detail would be missed using Boolean methods," adds Redgrave.

Whether advanced technologies such as Hapax’ are successfully applied to the wider market remains to be seen. If they are, perhaps the first thing many of us will notice is an absence of extraneous articles in our regular online reads. Whether or not this represents progress depends on individual surfing habits.