Replies to This Discussion

Permalink Reply by timothy vogel on October 31, 2009 at 2:02pm

interesting. 50 terms that mean essentially the same thing; Vincent is interested in using computers to analyze data. Wow. What a "Discovery"!

Does anyone here have any faith that any of these keywords, when used to ply Vincent with a host of data mining content (or golf clubs, books, or disability insurance for that matter) to "sample", that he will? Having been an architect and upper management in many "recommendation engine" applications from pre-dotcom, to dotcom1.0, through dotcom2.0 I will guarantee you that none of these keywords will generate any CPA metrics that differ by a significant difference.

The essence of keyword performance is generally measured in the tenths of percentage points, hardly more. Thus, true excellence in "content targeting fulfillment" will generate increases in orders of magnitude, not mere noise in a low-level signal.

To me, text mining to find keywords for online auctions is SO passé now. User Context is the upstream intelligence one can use to generate these huge leaps in performance, and even so that very pocket of performance will have an arbitrage advantage in price and not any effect on the overall fulfillment rate.

In the end, if you seek to drive overall CPC or CPA increases without ever trying to understand and accurately predict the consumer's visit intent, you'll merely be adding to the Google/Yahoo coffers, and certainly not those of the content providers themselves.

Factor in click fraud of the magnitude the Google/Yahoo's of the world simply cannot afford to admit exists or remedy directly themselves and you have "more of same". Don't forget that Google's "big solution" to clickfraud is moving from CPC to CPA, a move I think affords them no penalty and the bonus of knowing what we did once off their site that they never knew before.

Eric Schmidt is one financially crafty dude!

Keywords only work in an online environment designed by Google, in an auction setting where the price is more often set by the topmost desperate "content providers" who haven't discovered, as others there before them have, that those particular keywords won't pay off!

dotcom3.0? What will it hold? let's hope it is far less platform and far more true leverage of what we've learned about the consumers of our content. In a very real way, keywords are an easy, if not at all cheap, way to avoid understanding your customers and reducing their motivations to a series of clicks as opposed to a fully stochastic model of them as content consumers!

TV