Loading ...« Weekly Wrap Up - 9 May 2008 | Home | New Google Local Strategies »
Slow Down, Important Information Ahead
By Cameron Ferroni | May 19, 2008
Ah, yield vs. relevancy - topics that get batted around in search circles all the time. As part of our ongoing primer in the search space, I felt like today was a good time to tackle the definitions of each of these. Let’s start with relevancy.
We’ve talked a little bit in previous columns about various facets of relevancy. Today, I want to poke around at all of the facets. According to Webster, relevance is defined as “1 a: relation to the matter at hand b: practical and especially social applicability; 2: the ability (as of an information retrieval system) to retrieve material that satisfies the needs of the user.” Well, that seems pretty relevant doesn’t it! In fact, in the Search Advertising space, we focus in general on the relevance of an ad, and it generally applies in a number of different facets:
- User Relevance - Is the advertisement compelling to a user? Is there something about it that makes them want to click on the ad? In general search engines have rules around their ads to try to enforce this - they prohibit things that will turn users away, like ALL CAPS, or too! much! punctuation!, or excessive uses of modifiers like GREAT, or BEST, or fancy manipulation of text as in L@@K!
- Keyword Relevance - Is the ad reasonably related to the keywords that the user entered? Seeing ads for medicine when I’m searching for a restaurant as an example isn’t going to be very relevant. But would ads for a kitchen store be relevant? That’s a good question.
- Geographical Relevance - Is the ad targeted at a location that is relevant to the user - either where they are located, or where they are looking?
- Destination Relevance - Some search engines actually take the time to review the content of the page that the ad is pointing at to make sure that the page is relevant to the geography, keywords, and ad copy!
- Contextual Relevance - And some search engines go even further to look at additional refinements, like what the content on the page that the ad is displayed on to make sure that the ads are relevant, in particular when there are ambiguous keywords. Consider for example a search for “seal” on a music site - there it makes sense to show ads about Seal the musician - used CDs, upcoming concerts etc. On a search for “seal” on a Natural Geographic website, then perhaps it would be more relevant to show ads for nature shows, or for the local zoo!
- 2nd Order Relevance - This is my assertion that direct relevance is only one part of a potentially larger opportunity that the industry is currently missing out on. Namely that by showing ads that are related (movie ads when looking for a restaurant) we might be able to extend our yield!
So that’s relevance - what then is yield? Back to Webster - Yield has many definitions for yield, but the most relevant are defined as “4 a: to bear or bring forth as a natural product especially as a result of cultivation <the tree always yields good fruit> b: to produce or furnish as return <this soil should yield good crops> c (1): to produce as return from an expenditure or investment : furnish as profit or interest <a bond that yields 12 percent> (2): to produce as revenue : bring in <the tax is expected to yield millions.” In our sense, we are hoping to maximize revenue by focusing on the yield associated with our advertising. In other words, the more people that click on an ad, the more money that a search engine can make. In general, when we talk about yield we look at trying to maximize revenue by displaying the highest priced, most relevant ad for a given search.
So when you reflect on these two definitions, you can see how they are very intertwined, and how complicated it is to really solve this multivariable maximization equation - particularly for the computer. In an ideal world, you would have an ad that solved all of the relevancy constiuents, and was the highest bid upon ad in the network. But of course, it is never that easy. And that is where the magic comes in in solving yield. Maybe the ad is the most relevant, but has a slightly lower bid price - should you move that to the top anyway? Probably - but only if that ad gets enough clicks to warrant moving the higher priced one down. You also have to think about budgets - maybe the lower priced ad has a higher daily budget, so you want to get it more traffic, to maximize the budget - so go ahead and move it up. Maybe your relevance algorithms are wrong - or maybe they need to be tweaked to weight destination relevance over user relevance. It’s a highly complicated, multi-variable equation - but one that we need to be solving - not just to increase revenue, but to do the best job connecting users and advertisers!
Topics: Local Search, PPClick |
June 9th, 2008 at 8:14 am
Great breakdown of these two search algorithm factors, Cam - especially relevance. Your “2nd order relevance” point is interesting…while it seems like something we’d want to think about incorporating on sites like Openlist.com, think GYM will incorporate on their sites anytime in the near future?