Adsonomies

(Or why Yahoo needs to put me on retainer, Part Duex)

I will caveat this whole post by taking my ego off the table as my opening ante. For all I know, the good folks at Yahoo are already working on this, have built it, are ready to go to beta, are projecting record revenue, etc.. With that said, I have not heard about it yet so lets continue.

Some time ago I wrote a post about Yahoo’s potential (perhaps unintentional) brilliance in buying del.icio.us You can read the whole post here, but the simple summary is a follows. Del.icio.us is now widely regarded as an amazing source of classification of web pages. Using the concept of folksonomies (user generated taxonomies – or classifications), del.icio.us has instigated the largest open source project on the net: categorizing the content of web pages. It’s sort of a reverse Wikipedia - to some extent taking advantage of the fact that a lot of content is about more than one thing or is specific angle on something (which Wikipedia by definition is not). When I want to learn about SAP transports, I don’t go to Wikipedia and search on “SAP transport”, I go to del.icio.us and see who has tagged web pages with “SAP” and “transports”. I can find some pretty good resources specifically when I am looking for targeted concepts. If you run this thought out a little you’ll see that Yahoo, perhaps without knowing it, bought (for a mere 40MM dollars) a whole new way of letting people search the vast expanse of the web. Rather than Google’s reference based page rank or the more common yet complicated lexical analysis done by most search sites (e.g. figuring out what a page is about by looking the content in it), Yahoo could simply match del.icio.us tags (classification) to search terms and introduce a whole new concept of searching. Granted you can do this yourself by going to del.icio.us and typing in search terms, but my point in that post was for Yahoo to integrate this information directly into search for the 90% of people that don’t yet know what del.icio.us is but still want to find party pictures of Paris Hilton’s private bits. In doing this, they could possibly siphon off 5-10% of the searches from Google by getting people to think of tag searching before they think of Googling something. Cracking idea Robertson!

Today I had a new epiphany around this concept. It all started a couple of days ago when I became interim VP or Marketing for Newmerix. This in and of itself is a long story but suffice to say I am now actually responsible for rocking our marketing out and not just talking about rocking it out. One of the first places I have focused my time is on what I call “self-service marketing”. Its all the activities that you set up and they just sort of run for you, letting people interested in PeopleSoft and SAP change management opt into our lead generation process. You could include our web site (passive marketing) as well as search engine optimization and marketing. And since I met Dennis Yu, I’m totally obsessed with the later.

Dennis was introduced to me by David Cohen (of Colorado Tech Stars and other claims to fame). Dennis used to run Yahoo’s Internet Marketing program. That means that Dennis got to play with Google AdWords and Yahoo Search Marketing with a multi-million dollar budget and Yahoo’s home page at his disposal. Nice toys! Let’s just say that Dennis has forgotten more about keyword and content network marketing than most people learn in a lifetime. And, to our Colorado entrepreneurial community’s benefit he also decided to moved here to Boulder recently. With a little help from Dennis I was able to increase our click through rate by 20x in 2 days without spending one additional dime.

One of the first things that Dennis recommended was that we shift our dollars heavily towards keyword based impressions and less towards the content network. For those that don’t know much about SEM (like me!) there are basically two places you can stick an ad. The most familiar would be on Google’s home page. When you type in “SAP transport management” you should see on the top or side a little text based ad from Newmerix enticing you to come to our site, say hello, download a whitepaper and in the process opt into our marketing program (standard tech startup “give us your info and we’ll give you some content of value”). There’s no real magic here other than knowing how to turn the nobs and experiment and run through all the permutations of campaigns, ad groups, keywords, bidding, ad variations, rotations, etc.. (piece of cake huh). For those that are obsessed with day trading, the high is sort of similar. Place some bets, watch the results, harvest, adjust. It’s very fulfilling as you get almost instantaneous feedback from Google (Yahoo is PAINFULLY slow to return results to you). The second type of impressions that you can place are those on what is referred to as the “content network”. Ever read a blog about tennis and seen an ad for “Wilson Tennis Rackets” on the side of the blog. This is content network advertising. Google matches the nature of the page (content) with the nature of your ad (content) and decides where to stick them. As this is a less exact science the click through rates are massively different (if you expect at least a 1% click through on a minimally well managed keyword campaign you can expect maybe 0.1% click through on the content network). That said, I am discovering that the content network actually has some value. As a side note, ads on keyword impressions that perform poorly might actually perform really well on the content network. A good example is an ad impression we have entitled “PeopleSoft Change Management”. It gets almost no click through on keyword impressions (even when someone types “PeopleSoft change management” directly into Google) but for some reason performs really well on the content network. This is fundamentally why SEM is addictive – it’s a constant minute by minute game and optimization.

Okay – lets get back to Yahoo. Well, let me start by saying that the results you get on Google and Yahoo can be vastly different. Keep in mind that I am marketing for PeopleSoft and SAP change management and not ring tones, so this comment will clearly not apply across the board. This has to do with the type of user that searches each site. Yahoo seems to be heavily consumer focused (ring tones) and Google captures the business audience much better. My results are something like 50x better on Google (keyword and content network) with exactly the same campaigns, ad groups, keywords, and ad variations. In other words, Google gets 5-10x more of my SEM budget on a daily basis than Yahoo does simply because the users are different.

That said, imagine if for a second there was a better way to KNOW (not guess) what a general web page was about (not a keyword search but something on the content network). Instead of matching lexical concepts between ads and web page content, if I could match some form of content network keywords to place my impressions I am sure that I could get a better click through rate than Google’s 0.1%. Well, and it should be no surprise where I am going with this, Yahoo actually has these keywords. They are called del.icio.us tags. The del.icio.us users have already done the work for Yahoo. If there is a web page about SAP transports, chances are some del.icio.us user has tagged it that way. Well imagine if that web page is also part of Yahoo’s content network (e.g. serves Yahoo ads up on that page). Why not let me do keyword matches against del.icio.us tags. So Yahoo, without any more fanfare I give you my best idea of the day – the third type of campaign network – the tag network (or adsonomy). Set it up just like a keyword campaign but cross reference del.icio.us tags with content partners and search keywords and viola. I would bet that the results would at least approach the half way mark between a keyword based click through rate and a content network click through rate. And, it might actually give Yahoo a leg up on Google (or at least an even footing) which it looks like they seriously need.

That’s now two game changing ideas for Yahoo - all for free. The third is gonna cost them.

Comments

[David Cohen]

Interesting thoughts Niel. Glad I facilitate getting two really smart dudes in a room and talking. The power of that never ceases to amaze me. Told ya you'd like Dennis. ;-)

[David Henderson]

Mmmmm folksonomy based content targeting.... interesting!

My guess is it would perform about as good as contextual ads. It's just people (folksonomy) tagged content rather than Google algorithmically (contextually) tagged content.

The problem with content targeting is that it is about the context of the page/content not the context of the user. That's why it gets a 0.1% clickthrough rate.... or another way to view it is it fails 99.9% of the time.

Now delicious does have a lot of context about its users. Like what pages they think are important enough to tag, and when and what they tag them. Yahoo also has a lot more context about it's users across all Yahoo the services. Now we're talking! Yahoo could also expand it's user clickstream knowledge beyond Yahoo by enabling Web History (a la Google) in it's search tool bar.

This is where we're all headed here - attention recording - recently legitimized by Google. The big question in my mind is who's going to control the aggregation and use of this data? Yahoo? Google? Experian? Visa? Me.dium? or the user? Who's in control? Who should be in control?

Welcome to attention economy! ;-)

[Brendan Lally]

Nice matchup logic by delicous tags back to origonation site thats already "known"

If Yahoo (or others) could do better context extraction and categorisation (semantics web :) then that would help out a lot. Don't hold your breath too long.

Lal

[Stan James]

I love your concept of delicious as inverse-wikipedia. Del has a record of the *union* of every page that has *something* to do with a a topic, Wiki has a record of the *intersection* of all that data.

I fear the analogy breaks down, however, in that wikipedia is always editing itself and (mostly) spam resistant. The set of pages with a given tag cannot be edited, discussed or corrected. E.g. see the spam in this list:
http://del.icio.us/search/?fr=del_icio_us&p=viagra&type=all

Were delicious tags factored into primary search results, delicious would be spammers new #1 target.

Still, we at Lijit will be using delicious tags in quite a similar fashion, using the existing API to get the tags for a given page. E.g.
http://del.icio.us/rss/url/465d42f5164aaab15d6bac9f3e24dbd9

The whole thing recalls to mind Greg Yardleys post when delicious was purchased: "Do they have the cojones to use the tagging within del.icio.us - a pretty good measure of interest, importance, and subject matter - to modify their main search results?" (http://www.yardley.ca/blog/index.php/archives/2005/12/09/meaning-of-the-delicious-acquisition/)

(Glad you're blogging again, Mr. Marketer!)