Personalisation systems are a great idea, but costly and difficult to implement

Personalisation systems are a great idea, but costly and difficult to implement. Liz Warren explains why success stories are...

Personalisation systems are a great idea, but costly and difficult to implement.Liz Warren explains why success stories are a-typical

One way in which the Internet differs from almost all other media is that it supports two-way dialogue between customer and corporation. Yet the Web is also a very impersonal medium: you may be getting hits on your site, but you don't know from whom, or whether they've visited before. That's why Internet consultants are so excited...

by personalisation technologies.

Based around techniques developed in the traditional direct marketing industry, personalisation allows you to recognise visitors with whom you've had previous dealings and treat them as individuals. This means you can make content more relevant to individual users, which should encourage greater loyalty to the site or increase sales.

Ken Burke, CEO of e-commerce site developers Multimedia Live, says that personalisation can be used throughout a site in order to lead customers towards purchase or other desired behaviour. In the business-to-business (B2B) world, for instance, it can be used to present information specific to trading partners, such as previously negotiated price lists or an order history. This can streamline the order process and allow 'information self-service' by customers, reducing the load on call centres. The same tactics can be used in the business-to-consumer (B2C) world, when customers are making frequent repeat purchases, as on supermarket sites.

Personalisation can also be used in systems aimed at employees - especially knowledge workers - to give them quick access to the information they use most frequently, with content they use less frequently buried more deeply. This kind of personalisation can be achieved using basic technical solutions, such as cookies, which allow you to recognise previous visitors by name and personalise very simple aspects of the site. However, Burke explains, a further three levels of sophistication exist.

The second level uses rules-based engines to assign customers into groups on the basis of data gathered during registration questionnaires. Each group can then be shown different content. A third level is collaborative filtering, which allocates users into groups according to their behaviour as shown through clickstream data. Finally, predictive marketing uses statistical models to estimate - based on past behaviour and business rules - which content or offer will most appeal to the user. Each time the system acquires a new piece of data about a user, it recalculates its predictions as to what that user might do. This allows users to be assigned to groups much more dynamically than with rules-based solutions or collaborative filtering.

In each case, points out Russell Francis, consulting services director at electronic CRM specialists Infogain, personalisation can be either explicit, with customers actively and willingly providing data about themselves, or implicit, with profile data gathered quietly behind the scenes and not necessarily with the knowledge of users.

The personal touch

Notable examples of successful personalisation systems include Amazon's recommendation system and a number of B2B supply chain extranets. However, Peter Matthews, managing director of e-business consultancy Nucleus, claims these are the exceptions rather than the rule. "Personalisation is a great idea but difficult to implement; it's hard to justify the amount of money spent on it," he warns. "The idea that it's a panacea for online marketing is a delusion; successful examples are a-typical."

Burke confirms that experience over a longer period in the US has shown that personalisation can only increase sales by a few percent, falling far short of the extravagant claims made by some personalisation software vendors. For smaller companies, personalisation doesn't provide an adequate return on investment.

This is partly due to the high costs of implementing the more sophisticated forms of personalisation. According to Francis, you can spend about £250,000 on software and as much again on implementation services - and the aid of a specialist consultancy is almost mandatory. Even though the tools are, in effect, shrinkwrapped they require customisation, as creating the necessary data models and business logic is complex and challenging.

Once you've got your personalisation solution up and running, you may want to outsource its ongoing operation. While this is possible, warns Chris Trott, business development manager at data analysis tools provider Hyperion, "personalisation is part of the business process, so if you're going to get best value, you need someone who understands your business, your clients and what you're trying to achieve - and you can't outsource that."

And while personalisation can be retrofitted to an existing site, it's not a bolt-on: it needs to be closely integrated with the rest of the technical solution and the site business model. "We would strongly advise that you plan it upfront because it can affect the design of the website in terms of the kind of information you want to collect," Trott says.

Potential Pitfalls

However, the poor success rates for personalisation projects are not just the fault of implementation headaches but result from the potential weaknesses of personalisation itself. First, Francis points out, personalisation solutions can make rash generalisations and poor inferences, especially from clickstream data. For example, it may interpret the purchase of a Barbie doll as an interest in Barbie dolls rather than an interest in finding a present for a nine-year-old girl.

A second danger is overspecialisation, so that users miss out on some of the site's offerings or the wider range of information available from a portal.

A third factor, Trott points out, is that while many personalisation solutions are aimed at websites, you need to be able to draw in data from contacts made through other channels in order to get an accurate profile. You also need to be able to deliver the same degree of personalisation in your call centre or retail outlets as you offer on the site.

Furthermore, users often view personalisation with some distrust. "Customers don't necessarily want cuddly relationships with big companies because they often feel many big companies exploit their customers," explains Matthews. In particular, he warns, customers can be scared away if you are overfamiliar the second time you visit the site or if you bombard them with email campaigns that are poorly targeted.

Finally, Burke warns, it will take at least six months before you see significant results from a personalisation solution, because you have to collect sufficient data on customers and campaigns to allow you to start targeting offers. Rules-based solutions may start yielding results more quickly, but you still need some customer history against which to apply the rules.

Trott adds that personalisation technology will only deliver if you have the analytical tools and the marketing expertise to make sense of customer responses to campaigns. This way, you'll be able to refine your offerings and the way in which you target customers.

What is personalisation?

Personalisation technologies allow you to recognise customers, trading partners or employees and provide them with individually tailored content. Advocates of personalisation claim it can increase customer loyalty and boost sales, or help your own staff become more effective by giving them better access to key information. However, sophisticated personalisation solutions are costly and difficult to implement. In addition, it is easy to alienate customers if you classify their interests wrongly or appear to know too much about them.

Case study: WPA

Western Provident Association (WPA) provides private health insurance to UK organisations and individuals. Today, it covers 500,000 policyholders and has an annual income of £100,000m. Now, WPA has launched an extranet with personalisation features which is allowing individuals, corporate customers and brokers marketing WPA policies to manage their dealings with WPA.

Corporate benefit administrators are able to manage their entire benefit plan, analyse claims status, review premium payments, and view other information about employees' policies through a link on their own corporate intranet. Individual customers can use the solution to obtain quotes, buy policies, obtain pre-authorisations for procedures, submit claims online, and view the real-time status of open claims.

Finally, brokers can repackage the self-service features in WPA's solution and combine them with their own customised content and personalisation to deliver the same facilities to their own customers.

WPA's solution, which was developed by an in-house team of three in just six months, is based on products from Allaire which offer content management, rules-based personalisation and e-commerce.

"Certain policyholders require certain links and options that are relevant to their individual policies," explains Adrian Hodgkinson, WPA's head of e-commerce. WPA has been able to increase the number of policyholders overseen by each member of staff from 1,200 to 5,500.

Virgin Wines

Virgin Wines saw the Internet as a way to bring a wide selection of wines to consumers while offering the old-fashioned service which used to exist in independent high-street wine merchants, where staff would get to know the tastes of their regular customers and be able to recommend other wines they thought those customers would like.

To deliver this business model, Virgin Wines needed to be able to offer a one-to-one service based on an understanding of customers' tastes, either through analysing their previous purchases or by directly questioning customers. It chose a toolkit from Blue Martini which uses rules-based logic to show customers that are offers targeted at them. The solution also helps Virgin Wines bring people back to its site by using offline direct mail to present them with individually tailored offers of mixed cases of wine.

The range of functions within the Blue Martini toolkit allowed Virgin Wines' implementation partner, Accenture, formerly Andersen Consulting, to get the site up and running in just four months. However, Virgin Wines CEO Rowan Gormley warns: "It can cost as much in resources and energy to build good personalisation as it can to build a good website."

It's still too early to judge whether Virgin Wines will be a long-term success, but the Blue Martini solution is certainly allowing the company to recognise customers, speak to them offline and speak to them as individuals rather than just sending them blanket flyers.

It has also allowed the company to double its conversion rates since the site was first launched, simply by managing the store properly.

Questions to ask

  • Is personalisation important enough to my business model to go to the trouble of implementing it?
  • What kind of return on investment should I reasonably expect?
  • To which areas of my website should I apply personalisation?
  • Am I using the right software tools for the personalisation activities I want to undertake?
  • Does the design of my website support personalisation?
  • Do I have sufficient direct marketing expertise in my organisation to deliver personalised campaigns?
  • What are the potential pitfalls of my personalisation solution and how can I minimise the risk of these occurring?

    Personalisation suppliers

  • Broadvision -
  • ATG -
  • BEA -
  • Netperceptions -
  • Blue Martini -
  • Allaire -

  • Read more on IT jobs and recruitment