Peter Lawrence's The Making of a Fly is probably the most famous text of our time that uses science's knowledge of the invertebrate fruit fly to explain the larger concepts of developmental biology and molecular genetics. As you may have heard, at least a small part of its notoriety can be attributed to the fact that it was recently listed on Amazon for the princely sum of nearly $24 Million dollars. This example, though humorous, speaks to the potential hazards of allowing algorithms, particularly poor ones, to run with insufficient human oversight, a problem that can affect both the competitive price setting for books and the bidding of paid search ads. Michael Eisen, an evolutionary biologist looking to purchase The Making of a Fly initially found it priced at the bargain amount of $1.7M. After some nifty sleuthing, he was able to determine the algorithmic settings that were pushing the price to stratospheric levels. It turns out that two booksellers were adjusting their prices for the book based on what the other was doing. When the seller, bordeebook, raised its price, another seller, profnath, would raise its price to 0.9983 times bordeebook's price, prompting bordeebook to raise its price to 1.270589 times profnath's price (you have to love the precision here). And so on and so on until the price hit $23,698,655.93, at which point a human apparently and finally noticed and the price was dropped. long argued against position bidding largely because, as Google has confirmed, traffic value doesn't vary by position, we do receive requests from clients to hit a minimum ad position or to "own" a particular keyword. From a branding perspective this can be a reasonable strategy, so we are happy to oblige under the understanding that our strategy will be different from the typical ROI targeted approach. An obvious safeguard the booksellers missed was to have a sanity check in place on their algorithms, say one that stopped increasing the price when it hit the GDP of a small island nation. Fortunately for them in this case, they didn't have much to lose other than the chance to sell the book at a price someone was willing and able to pay. Had their algorithms gone the other way, however, they may have lost their shirts selling books for pennies or, more likely, faced customer ill will for canceling under-priced orders. In paid search, the consequences for runaway algorithms can be much more severe. Google isn't going to offer refunds for your bidding mistakes, so you'll be on the hook for your spend with little recourse. I would hope and expect that our friends in the industry have some notion of a sanity check in place for bids as RKG does, and further, smartly set campaign budgets to serve as guard-rails for unusual overspending. Still, even if you're avoiding a catastrophic scenario with those simple solutions, over the long term your PPC program could still be bleeding money due to an imperfect algorithm with a lack of supervision, and a slow leak can be harder to detect and ultimately cost more. Account reps and analysts that are actively monitoring performance and digging into the data will find these issues long before the bidding programmers. If your paid search analyst actually understands how his or her bidding platform works at a deep level and has the flexibility to apply client-specific safeguards and tweaks proactively, all the better.
Join the Discussion