On MaxBid CPCs
Amusing post over at the Google Adwords blog.
To whit: an Adwords advertiser set a max CPC at $5.00 for the phrase “flower delivery”, and then was shocked (shocked!) when Google sold him (her?) clicks at $5 a pop.
Google’s response is also amusing: “it’s always advisable to set your Max CPC no higher than you are actually comfortable paying for a single click”.
Three observations and one digression.
Observation One: It is easy to do PPC bidding poorly. It is very hard to do PPC bidding really well. Bidding is more than just picking a number. For direct marketing ads, the key is estimating the expected upcoming sales for each click purchased on that term, as a function of term, landing page, copy, ad position, time of day, day of week, and season. For brand marketing ads, the key is understanding the tradeoffs of impressions, clicks, cost, and position. For both, you’ve also got to keep an eye on your competition and the bid landscape. Smart bidding requires smart humans and great technology, both solidly grounded in real statistics and optimization.
Digression on Search Math: RKG has formal research collaborations underway with some great profs at Rochester, UVA, and UC Santa Clara. When introduced to this problem, academics invariably say something like “wow — bidding’s a really hard problem, much harder than we thought!” and start throwing out concepts like “non-homoegeneous non-stationary“, “game theoretic”, , “bounded rationality”, “stochastic optimization”, and “NP hard“. Yep. We know. Smart bidding algorithms are hard.
Observation Two. The maxbid-vs.-value curve can have a very flat plateau. That’s likely not the issue here for the hapless flower advertiser quoted, but this plateau is a deep and important issue for all advertisers (as well as for all search enines’ top lines). More on the maxbid-vs.-value curve in an upcoming post.
Overservation Three. The fact I personally find that Adwords blog post funny, though I don’t think that’s how Google intended it. Ack — I’m a search geek. :)