Editor’s note: Hank Nothhaft is the co-founder and chief product officer of Trapit, a personalized content discovery platform currently in beta. Trapit was incubated at SRI and the CALO project.
eBay’s recent acquisition of the recommendation service Hunch was an important score for the online retailer, giving it a way to mine the ever-mounting mounds of structured and unstructured data for more relevant and accurate consumer recommendations.
Such recommendation engines are already (and have long been) a necessity, not only for retailers, but for the entire Web. Every major internet company, from media outlets to social networks to software applications, is having to meet an expectation of better understanding their customers as individuals, to provide them with information and suggestions that they themselves may not even have realized they want or need.
Integrating better recommendation algorithms and services was really just the first part of a larger, necessary movement to make the Web more personalized. As we watch the ongoing struggles of search engines to provide relevant yet deep-diving results, or Facebook’s fruitless attempts to better identify which content shared by your friends is most important for you to see, it’s clear that we need something smarter, something more sophisticated than mere recommendations and customization. Personalizing the Web is one of the most important and difficult engineering tasks we now face in the evolution of the Internet.
Recommendations Have Hit the Ceiling
Amazon and Netflix once stood as exemplars of recommendation, providing suggestions based on what other users with similar habits and product histories selected, with a touch of genetic genre data thrown in. Yet both have faltered in recommendation relevance as the crowd-sourced approach has become more of an echo chamber than a personalized filter. (And Amazon certainly does itself no favors with its barrage of “recommendations” for its own Kindle Fire).
These recommendation engines were once ground-breaking, but they have failed to evolve. And more importantly, our expectations as Web consumers have evolved beyond the simple concepts of “users who purchased item X also purchased item Y.” At best, services that claim personalization based upon these aggregate metrics attempt to triangulate an identity for us as individuals based upon the galaxy of other individuals. They try to pin us down into an archetype, into a box of likes and interests, without recognizing that as humans, what we desire, want and need is in constant flux and ever-evolving.
In all fairness, that’s an incredibly difficult awareness to achieve. But its disingenuous to attribute “personalization” to services that are really just crowd sourced general interest mapping. And those results are just insultingly banal!
Getting Closer, All the Time
The acquisition of Hunch will hopefully signal a shift from recommendations to actual personalization. This is not because Hunch is such a precise personalization tool, but rather because it is an excellent recommendation engine. As more relevant recommendations become ubiquitous as the standard across the Web, we can finally begin to aim to shoot beyond that baseline to realizing a personalized Web.
Over the next few years, technology that truly understands and recognizes us as dynamic and unique individuals rather than types will be the predominate trend. Siri’s adaptability and cognizance is the first major step in this direction, and we will begin to see that type of finely-tuned, perpetually trained artificial intelligence helping turn the Web and our technology more and more towards us as individuals.
It’s a movement from the Web mentality of searching, to one of delivery. It’s a shift from pull to push technology–and it’s happening because there is enormous upside for the first movers in the era of personalization.
The shift to a more personalized Web is all about revenue and customer/user experience.
Consider the shift that is still occurring with our televisions. Whereas we have long considered our television sets to be mere boxes for broadcasting pre determined content to us as passive consumers, we are increasingly taking control of that content. This extends beyond mere DVR capabilities, as Web-enabled televisions begin to offer a new layer of personalization.
Our televisions will eventually become truly our own, unique televisions. Better yet, we will have accounts that we can log into that will personalize any television set with programming specifically selected by and for us.
Groupon, no doubt, is itching to find ways to personalize each and every offer they have for each individual it is sent to. Groupon knows that targeting by regions increases conversion and sales, but imagine how much they could amplify that effect if they were targeting based on a rich and sophisticated understanding of the individual person that receives each offer?
We can imagine this type of personalization applied to all of our current technologies, but it first requires a fundamental, philosophical shift in how we think about and understand the notion of personalization. A big part of that means that recommendation technology needs to be properly understood as the tip of the iceberg — it’s table stakes and nothing more. The real game is played with true personalization and a sophisticated understanding of individuals and all of their unique, ever-shifting personas.
The Endless March of Personalization Progress
At this stage, personalization is best achieved as a mashup of our interest graph, social graph, individual input, and Pandora-eque qualifications of structured data. When maximized, this can work quite well, but we can’t stop there.
As the Web becomes our own personal web, the technology needs to register and understand our flux in personality. This means incorporating both more direct and more ambient information, such as awareness of time, location, my schedule, my habits and engagement with content. Furthermore, it means realizing that human identity is a constantly shifting target. The work of personalization is never done, even if it is done with less direct input and feedback from me personally.
Content creators, marketers, sales professionals and publishers crave that myth of stability in defining their users and audience. Yet as the Web has shifted to become dominated by the stream metaphor, that myth has been easily eroded.
This is not a bad thing, even if it makes understanding that customer baseline upon which we build businesses that much more difficult to determine. It was always a myth to begin with — we simply did not have the data or the tools to operate otherwise and recognize users as individuals, as the amalgam of ever-shifting interests and personalities that they truly are.
What replaces that baseline of stability, however, is the flux of new opportunities, of understanding our customers and audience in increasingly focused and nuanced terms. Likewise, we as Web consumers are starting to expect this understanding from the websites and businesses with which we choose to interact.
For companies, recommendation is the gateway into recognizing the value of a much more attuned personalization. And as recommendation layers become ubiquitous, we can now finally begin moving beyond it, to achieve personalization as the next great triumph of the Web.