Enter your email address:

Delivered by FeedBurner

About This Blog

Josh’s Tweet Stream

  • More tweets

« Ask Jeeves buys Bloglines | Main | NYTimes.com purchase of About.com –it’s all about the people »

February 16, 2005

The slippery slope of search semantics

By Charlene Li

I’m embarking on a new research project, which involves sizing online advertising’s growth over the next five years. One of the issues I’ve come across is trying to define and size search marketing.

Here’s the problem – it’s pretty clear what “real” search marketing is – it’s putting text ads next to search results on sites like Google, Yahoo, MSN, et. al. The next variation is putting text ads next to search results on non-search sites like the nytimes.com. It’s still ads around search results, just not necessarily on search “sites”.

Now it gets trickier with contextual ads. The marketing logic is that someone who types in a keyword(s) in a search box is expressing a specific intent to find content on a particular subject. That intent is implied when looking at an actual content piece, hence, contextual advertising where the keyword is inferred from the content, and purchased explicitly by the advertiser.

Is this search? Contextual ad providers are typically also keyword search providers (think Google, Overture, Kanoodle). But this is more an extension of an existing marketer relationship than of the actual “search” product.

Another variation and extension – what if that intent is expressed as behavior rather than content? This has led to the birth of contextual advertising, offered via providers like Tacoda and Revenue Science on content sites like weather.com – and provided on a grander scale by Yahoo via their Fusion Targeting program. In most cases, these ads have been sold as graphical display ads targeting users who are designed “interested car buyers”, for example. But they are increasingly offered also as text ads, again, targeted to a specific group of users based on their intent as expressed in behavior.

So can we just call this space “sponsored links” and be done with it? Not so fast -- let’s go even further now with the image ad alternative that Google offers as part of AdSense for Content. Marketers have the option of showing an image ad instead of a text link on these content pages. It’s a thin line between these image ads that may appear on the side and bottom of the content, and the high-priced, CPM-based display ads that appear at the top of the page. In fact, doesn’t it make sense in the future for Yahoo! to extend its Content Match capabilities and place display ads contextually throughout the Yahoo! Network sites?

I think you begin to understand why I see such a slippery slope – once you take search though its natural evolution, it becomes hard to separate and size where search beings and traditional display online ads begin.

Now here’s the rub – I believe that in five years’ time (the scope of my sizing report) that marketers will be buying search and display ads in very similar ways. Marketers will be able to target display ads based on a combination of audience demographics and implied, contextual keywords. They will also be able to buy text ads that appear next to search results based on a combination of keyword search but targeted with different creative for men versus women, or customers versus non-customers.

So, what’s really “search” in 2010? How do you think we’ll be defining search in five years? Share your thoughts in comments below, or if you’re feeling shy or proprietary, send me an email.


TrackBack URL for this entry:

Listed below are links to weblogs that reference The slippery slope of search semantics :

» Charlene Li's Blog: The slippery slope of search semantics from del.icio.us WebCites
Search does not have to be just typing into a search box. These are really ads based on information consumption patterns. Search just happens to be the most widely-known version of that.... [Read More]


Greg Linden

In terms of advertising, I think you might see a model where advertisers throw their messages into a large pool with hints on the target audience. An automated process would test, find receptive audiences, and optimize the targeting. This is Google AdWords on steroids, where the keywords given are merely hints to an engine that experiments, learns, and finds the best targets for your ads. John Battelle has suggested models along these lines a few times, most recently on this post (http://battellemedia.com/archives/001262.php).

In terms of search, I think personalized search is inevitable. In the quest for greater relevance, the history of what individual users have done is just too useful to ignore. But it is a hard problem to solve well. Current search engines rely on having a single relevance rank for their speed and performance; introducing millions of relevance ranks, one for each user, creates a challenging problem.

Christopher Rines

Greg Linden makes some stellar points!

I think if we look x (no I am not sure what x is at this point) months/years out we will find a way to automate the delivery of a "personalized" advertisments to users in an automated fashion which would in effect be a one-to-one relationship. Once we are there it really doesn't matter if an advertiser is paying for a listing or it's a "plain" search result as long as the information is correct.

Personally I don't care if an advertiser puts a paid text link in front of me IF it is relevant (best case exactly) to what I want.

If we look at the yellow pages as an model/example we see only paid for listings but they are always relevant to what I am looking for.

That's all I ask for...

In a nutshell search is/should really be about finding the 1 link I actually want/need.

Christopher Carfi

One scenario is that, five years from now, "search" is so ubiquitous that it just *is* the way we interact with information. There may be no "search." It'll just be part of the interface. No one talks about "GUI v. text" interfaces anymore. Same thing.

Search could completely change the way we interact with local information, per here:

Addionally, think about the way the Gmail interface has completely changed the interaction mode with email. Navigating email used to be "sort by name" or "sort by date" or "sort by folder." Now it's "remember a couple of things about a message...sender, or a comment, maybe, and put it into the box."

So, what’s really “search” in 2010? One scenario says "it's the interface."

Jason Ratters

Interesting article Charlene.
If you are trying to size online ad growth I think it is better to focus the question on how much performance based or CPC advertising there will be -- not whether it is search or non-search advertising. The interesting dynamic over the next few years will be the extent to which traditional (non-performance) online advertising migrates to PPC-based. This is going to happen on search and non-search sites. Today PPC advertising on non-search sites, in terms of total ad revenue is probably < 5%, this could easily be 50% in 3-5 years. Meaning that 50% of a publishers online ad revenues will be performance based.

The next question is whether the migration from non-performance online advertising to performance online advertising will grow the online ad market, keep it flat, or even shrink it.

Then we need to consider the migration from print to online ...

Tony Gentile

Charlene... Unfortunately, no trackback... some thoughts for you here:


Dave Chase

In response to your query, "what’s really “search” in 2010?", take a look at this recent 3-part article "2010 Predictions" -- http://www.imediaconnection.com/content/5086.asp that took up the issue of where Google's growth will take it and laid out a fictitious scenario from a consumer's perspective of what combining "search marketing", behavioral targeting, inventory yield management and TV advertising would bring. Charlene is justified in laying out the issues of trying to capture "search marketing" as a discrete category when the lines are increasingly gray.

Danny Sullivan

"It's pretty clear what 'real' search marketing is – it's putting text ads next to search results on sites like Google, Yahoo, MSN, et. al."

No, that's search advertising. Search marketing is the combination of getting better listings for free and doing advertising. Thinking PR + Ads in the real world. If you only do ads in the real world, that's not the entirety of how you can market.

As for contextual ads, they are not and never had been search in my book. If you don't perform a search -- if you aren't in searching mode -- how is that search? The search companies like to pitch contextual as search because it makes the selling easier -- but it's not search simply because they sell it.

So what's a search ad in 2010? Something that was triggered by an actual search intent. You did some type of search activity and got an ad -- be it an image ad, an image ad on another site delivered based on past search behavior or a search ad showing up after you performed a keyword search somewhere.

Not in search mode? Simply reading something and you get an ad based on the content your are reading or the non-search surfing behavior you've done. You didn't search -- you weren't in search mode and that's not search advertising.

Ian Saunders

Danny is absolutely right. That contextual advertising is being aligned with search currently is only a consequence of the fact thate the leading exponents just happen to be search engine companies. In my opinion, contextual advertising should be ubiquitous. If as an industry clickthrough rates are to rise then it must be the case that all ad content, and by that I mean banners, rich media etc, should be relevant to the context of the webpage and well before 2010.

What is clear is that the techniques of contextual advertising must improve as the cracks are appearing in the contextual advertising veneer. The reason is that current solutions are largely based upon individual keywords. Such systems simply identify a keyword and present an advert based on that keyword. In a great many cases, the context of the advert is far removed from the context of the web page, which is why we get absurd ads such as those for white slaves for sale on eBay on Google.

Quite obviously, there is a need to be smarter. In order to distinguish the different senses in such examples as the above, we need to identify exactly which associated words in the sites and ads relate to the required meaning. Processing the data using a simple statistical metric - such as looking for the most frequently occurring words - has not worked. Identifying relevant context is something the human brain does instantly and intuitively; but capturing this intuition, and turning it into a workable procedure, is a very different matter. All ambiguities have to be anticipated. It is not, after all, just a matter of a two-way choice. There are half-a-dozen commonly used senses of 'bridge' in English, and if a site or ad wants to focus on one of these, interference from the others has to be eliminated.

Plainly a radically different approach is needed. My own company, Crystal Semantics has developed a linguistics-based 'sense engine' which analyses the lexical content of advertisements and links them with relevant Web pages, and also analyses the content of Web pages to ensure relevant advertisements.

Charlene Li

Thanks to everyone for all of the comments -- I really appreciate it. I especially liked Tony Gentile's lighthearted referral to the Pit of Despair from one of my favorite movies, The Princess Bride.

Dennis Caffey

Contextual Ad delivery is really a "push" technology..
relationships between the searcher, what they are searching for, and where they are searching are all factored into how Contextual is delivered...

Amazon really leads the way when it comes to contextual delivery...they have really tweaked the user data (because they have access to very specific data) so they can present more buying options...this is the future of search in my opinion...not just the static algorithmic presentation of numbered results...but rather a whole sphere of options...(this will work very well in the B2B arena for value added services and products)

As more niche sectors break out and build vertical this type of search marketing will become more important

Yahoo also uses this approach for their registered user base...

Search Marketing is about getting your message in front of people searching for your product/service at the moment they are searching...be in from the SERPs or some sympathetic relationship

The comments to this entry are closed.