Nice Knowing You
- Article 1 of 6
- Database Marketing, March 2002
Robert Buckley outlines the challenges that face Web marketers in identifying and tracking customers and the current methods used to keep them in sight.
Page 1 | Page 2 | All 2 Pages
And DoubleClick should know. The company ran in trouble in 1998 when it merged with Abacus, a company that maintained information about customers of catalogue companies, and has just dropped a consumer tracking service from its US portfolio of services. “The media and privacy advocates just saw the merger, put two and two together and made five. They assumed that companies with two large datasets would merge those datasets and build up profiles of people and what they did online.” Because DoubleClick adverts appear on many sites yet come from its own servers, the company has been able to build up a database of sites that visitors have been to using the unique identifier in its cookies. It has 100 million anonymous profiles of browsers’ personal interests for tailored advertising - the service it has dropped because of the slowdown in online advertising, it claims. But, says Chandler, “to link to Abacus’ database, the databases needed to have something in common. And there was nothing common to the cookie clickstream data and the Abacus database.”
Without some way to match clickstream data to offline data, it is impossible to combine the two. Name and postcode from web site registration information can be used as the key fields for linking the two. Loyalty card numbers are another way for the card provider to link offline and online purchase histories. Using credit card data would breach contracts with credit card companies, however, so before combining any on- and offline datasets, “take good legal advice and make sure you don’t do anything that would shock the average customer”, says Woods.
Nick Bidmead, CEO of NCorp, whose Ijen technology profiles users via clickstream analysis, says that “a privacy statement on customer-facing propositions” helps people understand what is being done with data collected on the site. DoubleClick requires those who use its ads to have links in their privacy policies to its own policy for the same reason.
Woods points out that a web site is doing well to convert 3-4% of its visitors into buyers or to register, so that means 95-99% are anonymous visitors. Behavioural analysis avoids privacy concerns, is more productive than attempting to tie in with offline lifestyle datasets and “many people enjoy the anonymous surfing experience”. Both his and Bidmead’s technology observe the interactions with the site made by visitors using clickstream and search engine analysis. The Auto Trader site uses Ijen to understand what is important to individual customers using the searches the customers make and the results that interest them. “It might find people are price sensitive or they prefer prestige German cars, for instance.”
Mitcheson agrees. “You can pretty much do all the personalisation by group without knowing the identity of the individual and flying into data protection implications. And if you keep all the data, you’ll run in data storage issues which are horrendous. The vast majority of profiling isn’t one-to-one marketing: it’s done as quite tight segmenting, targeting groups of 10-20 or 100-500 rather than a million. The biggest uplift comes from getting the web infrastructure sorted by identifying how it’s being used, the sweet spots and which marketing campaigns are drawing in visitors. You can get average spend up by another 60%.”
“The difference between web sites and department stores,” says Thomas, “is that you can break up a web site and segment it in many more different ways than a department store.” Online customer analysis makes use of different techniques to its offline counterpart, but produces similar results. While traditional marketers may want to use postcode-based lifestyle data and other tools of the offline world for segmenting their online datasets, the greater numbers of interactions and tracking techniques, together with the anonymity of the web mean that behavioural profiling produces far better results without the privacy concerns.
Case Study: Motorola and Ogilvy Interactive
Motorola planned to run an online advertising campaign last year for various mobile phone handsets and wanted to be able to track how customers responded to the ads and to its e-commerce site. Ogilvy Interactive, which designed the campaign, used WebAbacus to track visitor behaviour.
“There was a different creative execution per model,” says Janet Winslade, project manager at Ogilvy Interactive. “WebAbacus tagged up the whole batch of sites created for the campaign and for each visitor, we were able to take the referring source, which creative they’d clicked on, the position on the site, and whether they’d gone through to purchase anything. We were able to give Motorola a picture of how they used the site.”
Motorola’s existing customer database was segmented by behavioural types and the company segmented the clickstream analysis by customer type. The company compared how it expected visitors to behave, based on its existing segmentation, with how the visitors actually behaved, using the visitors who went on to make purchases through the site to match datasets. “What they found was some really strong correlations between the expected user behaviour and the behaviour they found. But in selecting sites for the creatives, while there were some good matches, some weren’t.” The company was able to determine which creatives on which sites were bringing traffic to its own sites and adjust accordingly, reducing the distance from entry to the site to purchase at the same time.
In addition, Motorola now has a better idea of which sites are the best with which to partner for its developing affiliate campaign. “We could demonstrate the ROI all the way through,” says Winslade. “And from the point of view of advertising, the numbers were very good, quite encouraging. Although you hear a lot about how no-one clicks on banner ads, a good proportion do.”
Page 1 | Page 2 | All 2 Pages