Friday, April 11, 2008

Cheap personalization using the referrer

Danny Sullivan at Search Engine Land writes:
Previous Query refinement is now coming to unpaid or "organic" search results, [Google VP Marissa Mayer] said.

For example, if someone were to search for [spain] and then [travel] after that, BOTH the ads and the organic results will be altered to take the previous query into account. To some degree, it will be as if the second query was for [spain travel].

This is a big deal. Big deal. It means that the results for many "single word" queries, which can be hard for sites to rank for when billions of listings come back, will become queries involving two or more words -- and much more specific ones.
This is a cheap and easy form of personalized search and personalized advertising that uses the history stored in the referrer instead of maintaining a separate, longer history.

As Danny says, the goal of personalized search is to help us find what we need faster and more easily by creating a dialogue with our search engine. No longer is each search treated as independent, but what we find or do not find in earlier searches impacts our future searches.


Joe Lazarus said...

Niall Kennedy has an interesting trick related to this...

He sniffs visitors' browsing history through their browser and then personalizes the experience based on which websites the person visited recently.

Shirish said...

I assumed that such a thing (taking the previous query into account) was always present Google. Surprising to see that it is coming now.

Why do you say it is personalization? It is not taking a user's profile (which indicates the user's taste and preferences).. all it considers is the previous query.

Greg Linden said...

Hi, Shirish. It may depend on how you define personalization, but I consider anything that shows different content to different people based on their past behavior to be implicit personalization.

Although this is just using the last query, it is showing different search results to different people based on a small part of their history. So, I think it does qualify as a simple form of personalization.

Anonymous said...

Although this is just using the last query, it is showing different search results to different people based on a small part of their history. So, I think it does qualify as a simple form of personalization.

So, wait.. it sounds like you are claiming that, basically, "personalization" is a superset of interactive search (aka search as a dialogue, aka search "sessions").

I don't quite know if I would agree with that characterization. It feels like you are defining personalization a bit too broadly.

I mean, if that's the definition, then would you characterize (let's use my favorite example) Vivisimo as a personalization search engine? Because when I go on vivisimo and type "spain" and then click the "tourist" cluster, I get a different result that someone that typed "spain and then clicked the "government" cluster.

So you would call that "explicit" personalization, rather than "implicit" personalization? It's basically the same procedure as Google.. sequential queries, the 2nd of which is affected by the 1st.

Greg Linden said...

That's a great point, Jeremy.

So, yes, I do think personalization is a superset of search as a dialogue.

But, you make a good point that helping with search refinement doesn't seem much like personalization, even if the query suggestions are based on other people's behavior.

I don't really have a good answer for that other than to plead shades of gray on the definition. There are aspects of personalization in these features, but the experience clearly is not fully personalized.

What do you think, Jeremy?

Anonymous said...

My own personal opinion (stress: opinion) is that the sharpest conceptual dividing line is between implicit (aka unintentional, aka passive) and explicit (aka intentional, aka active) applications.

As far as I understand it, most of the personalization that you talk about resides is in the implicit (unintentional/passive) domain. Most of the tools stuff that I yammer about resides in the explicit (intentional/active) domain.

The way that I would tease out these shades of grey would be to distinguish between an algorithm, and the application of that algorithm.

The same underlying mathematics (algorithms) can be applied in different ways, to produce completely different outcomes.

Personalization, to me, is one possible application, i.e. application in the implicit/passive domain. Search as a dialogue, on the other hand, is another possible application. I tend to think of the latter as residing more in the explicit/active domain.

Anonymous said...

I just had another thought, a better way to express it than I did yesterday. And it relates to the better data vs. better algorithms theme that you post on, every once in a while.

See, whether a search interaction is implicit vs explicit (passive vs. active, etc) is not just a matter of HCI design. It is a fundamentally different type of data that you are collecting from the user. In one case, the data has strong intentionality behind it. In the other case, it doesn't.

And since data drives algorithms, the directions the algorithms take with one type of data are very different than with a different type of data.