Online personalisation and recommendation technology can begin to replicate the hand-selling approach delivered by shopkeepers a century ago. Alison Clements asks if it is getting results

Online personalisation and recommendation technology can begin to replicate the hand-selling approach delivered by shopkeepers a century ago

Does your website sell from the moment customers land there? Personalising the offer has been considered one of the best ways to drive web sales since Amazon’s ‘recommendations’ techniques grabbed our attention in the late 1990s.

Personal recommendations – ‘Customers who purchased X also purchased Y’ – was the starting point in this new marketing discipline, and lists of ‘current top sellers’ also began helping consumers sift through the overwhelming volume of product options at their fingertips. But much more can now be done by using customer data, items viewed, demographic data, personal interests and favourite brands, authors and artists to personalise the offer. Etailers are progressing beyond using other customers’ shopping habits, to using an individual’s online behaviour to shape what’s presented to them.

Amazon’s oft-cited success has come from recommendation algorithms used to personalise the online store for every customer. “The store radically changes based on customer interests, showing programming titles to a software engineer and baby toys to a new mother,” explains an Amazon spokesman. “The click-through and conversion rates — two important measures of web-based and email advertising effectiveness — vastly exceed those of untargeted content such as banner advertisements and top-seller lists.” An individual’s transactional history comes into play, but Amazon’s referencing tools also allow the site to automatically suggest books of similar subjects to the ones a
customer is looking at – viewing Nigella books might lead to Nigel Slater books or cookware items being suggested too, greatly increasing the possibility of impulse sales.

Personal touch

Email marketing shaped by customer data is getting ever more sophisticated with new developments such as Amazon.com users in the US being able to link their Facebook account to their Amazon account, as announced this July. The potential is there for Amazon to show these customers recommendations based on their Facebook interests and activity, making the marketing messages far more personal.

Other etailers are catching up with the web giant’s personalising prowess. Personalisation is activated with log-in triggers, or the use of cookies. So Karenmillen.com welcomes logged in customers by name, and Johnlewis.com visitors are instantly reminded of their ‘previously viewed items’ on the home page, for example.

Segmenting customer groups can be one starting point. B&Q has worked with ecommerce specialist ATG to tailor its online offer to three different online customer personas – homemakers, DIY-ers and trade professionals – and is now delivering cross-channel shopping experiences based on visitors’ preferences. “Because all our website visitors are different, we had to find a way to offer them an experience specific to their goals in logging onto our site,” says B&Q multichannel development manager Joanna Robb. “The ATG platform enables us to guide customers based on their shopping mission, steering them to content that is geared toward their particular interests,” she says. For example, a customer seeking design ideas will be taken down the ‘inspirational’ path, which displays photos, room sets and creative ideas for home projects before introducing items for sale. The tradesman persona is offered a ‘functional’ route, and is taken more directly to items through the use of a Quick Finder filtering tool.

The vital statistics

“Another fast-evolving area in recommendation is using data on customers’ purchases to not just offer more of the same products, but to plug into relationships between product groups,” says Darren Vengroff, chief scientist at provider of dynamic personalisation RichRelevance. Web stores can begin to recommend products associated with favourite movie stars or authors, for example, or tailor their fashion offer to the discovery that people buying Ugg-style boots are also interested in denim miniskirts. “Being wary that algorithms will only be relevant for one season in many cases, particularly in fashion, is important,” says Vengroff.

“There’s a danger now that consumers get annoyed with recommendations they don’t want”

Aurora group multichannel director Hash Ladha

At Aurora Fashions, investment in multichannel technology is driving a strategy to ensure customer satisfaction and a seamless brand experience at every touch-point, as well as pursuing the commercial goal of increasing sales. “So we are very conscious that recommendations must be relevant to customers’ needs and in keeping with the wider Coast, Warehouse, Oasis or Karen Millen brand experience, while still encouraging spend,” explains Aurora group multichannel director Hash Ladha. “There’s a danger now that consumers get annoyed with recommendations they don’t want, particularly in fashion where people have a very individual look. We believe the best way to meet expectations online today is to use customer data and CRM to a certain extent, in conjunction with the right degree of up-selling and cross-selling on what is actually being viewed.”

Avail Intelligence CEO and founder Pontus Kristiansson says just as shopper behaviour has been analysed for merchandise planning in physical stores for many years, live online behaviour can be the key to up-selling and cross-selling. Avail Intelligence specialises in collecting anonymous visitor behaviour patterns in real-time, analysing data on pages viewed, items selected, click-throughs and time spent on product details. “We then compare items being viewed with the collective intelligence from previous behaviours, and mathematical techniques can present to website browsers what is deemed to be most relevant to them,” says Kristiansson. “It’s a case of narrowing down a thousand SKUs to saying ‘we think you’re probably interested in these five products’. This starts to address the problem of online shoppers not being able to make up their mind, and abandoning the purchase.” Avail Intelligence’s personalisation and product recommendation engine is used by etailers including Game and La Redoute, with the aim of generating significant increases in sales per visitor. 

Key to success

Amazon famously built an in-house personalisation engine and has accrued years of expertise. Very few other players have the resources to fund a team of 10, 20 or 30 software engineers, and experts in the science of algorithms, designing and running an in-house recommendation suite. It also takes years of learning before reliable behavioural patterns emerge and start delivering a return on investment. For these reasons, many etailers are opting for add-on solutions from analytics vendors or ecommerce platform providers, or hosted, pay-as-you-go services that already have rich market intelligence to draw on.

“Most importantly, marketers and merchandisers must know exactly what they are trying to achieve – be that attracting new customers, up-selling, increasing conversions or driving loyalty − before investing in the analytical technology required for personalisation,” says Vengroff. “The most successful personalisation solutions employ a variety of recommendation strategies, and are continually improved and remodelled to incorporate changes in inventory data and user behaviour.”

Personalisation techniques

  • Shoppers setting up their personal profiles This will clarify the kinds of editorial features, products and services they want to see, and is activated every time a unique user logs in.
  • Collaborative Filtering This is the process etailers use to track customers’ likes and dislikes and look for similar patterns. It’s a way of looking at the behaviour of people who are alike, and then making a set of recommendations based on what a particular group of similar people think is popular.
  • Versioning Displaying different versions of pages based on segmented customers. For instance, customers who arrived through a search engine are in a different segment than customers coming specifically to the site.
  • Interactive Filtering Solutions Displaying specific information based on a customer’s choices when they are asked which ‘direction’ they want to go in.
  • Recent Behaviour Pattern Matching Analysing real-time user behaviour in order to choose products to recommend. As people browse they are giving clues about what they are likely to want to purchase.