Is recommendation the key to personalization?

At the end of 2007, Amdocs bought the Irish personalization software house, ChangingWorlds, for €47 million. At TM Forum’s Management World in Nice last month, talked to Michael Lurye, director, product marketing, Amdocs Interactive which incorporates ChangingWorlds, about the reasons why.
Operators have been talking about the importance of personalization for several years, so what’s the hurry now? The short answer is that the emergence of the iPhone and the flood of copycat app stores has proved their assumption right – people so want to personalize their devices and services they use.
Lurye explains, “It has always been an early adopter idea, but now it’s moving into mainstream thinking in the context of mobile and fixed web usage, and digital commerce.
“Operators see it as a key driver for acquisition of customers and all the Requests for Proposals (RFPs) we receive for digital commerce from large operators includes personalization requirements. We acquired ChangingWorlds because we recognised capitalizing on personalization is critical for all our customers.”
ChangingWorlds has 58 installations around the world, including O2’s portal in the UK (it has run the software since 2003), which gives customers access to content and services, such as a library of widgets from which they can create their own home page. The software also learns what sort of content visitors like, based on browsing habits and other information, and makes recommendations along the same lines as Amazon.
Lurye adds, “The trouble is operators tend to have lots and lots of data at random points, but no easy way to translate it into action points.
Amazon was at the cutting edge with its recommendation service and operators want to follow that example. Operators are primarily concentrating on selling digital content, but there’s no reason they can’t sell physical goods too, such as accessories, phones, bundles or ticket sales such as O2 providing tickets for concerts at its The O2 arena in London through a one click purchase.”
Indeed, recommendation seems to be at the heart of Amdocs’ notion of personalization via its ChangingWorlds acquisition, but recommendation itself is the subject of fierce debate.
Many, including the Wharton Business School academics Kartik Hosanagar and Dan Fleder argue that in fact recommendation – and Amazon is always the example cited – simply encourages consumers to focus on the most popular items available and puts them off exploring for themselves. They argue recommendation "reinforce the blockbuster nature of media." They warn that, by deploying standard designs, digital retailers may be recreating the very situation, limiting choice, that the recommendation lobby often claims it helps customers avoid.
The way around limiting choice is a recommender algorithm that offsets the effect of the bestseller bias, but this has given rise to further rows about the validity of business models used and how it is done – and with what success – depends on how well the algorithm serves the business model. Is the operator wanting to sell tickets for next week or shift a back music catalog?
It will also depend on having a wide-awake algorithm writer that understands business needs as well as the technology. As history has shown, in telecoms this is not a common phenomenon.

Recommendation aside, Lurye continues, “Further down the line, personalization will help operators build bigger picture to understand their customers’ behavior better – ChangingWorlds can aggregate subscriber behavior data and this is the next thing its 58 deployed customers are looking to do.”
What took them so long? A pretty big shift in approach and attitude, perpetuated by a combination of many massive failed back office transformation projects and a new urgency engendered by the economic meltdown for pragmatic approaches that have a pretty quick effect on the bottom line.
Or as Lurye puts it candidly, “It doesn’t require a massive data integration project to pull information from many sources. Yes, it’s nice to have more, but for instance, it is possible to target adverts effectively based on very limited demographic information and people’s previous responses to adverts. True the more information you have for profiling the better, but you can still be pretty accurate through a partial view of a customer from a ChangingWorlds-type view.”
On that note, I leave you with a comment from Jonathan MacDonald, veteran advisor to brands on their digital marketing and advertising strategy: “Service providers have got the psychology all wrong regarding monitoring behavior to serve targeted adverts. If I’m spending a lot of time on certain car sites, it’s just as likely I’m planning to sell a car as buy one, so sending me a targeted advert is completely irrelevant and very annoying. Inference based on assumption is the mother of all cock-ups.”
It will be interesting to see.