Databases Reference
In-Depth Information
Now, one may ask: “Why let advertisers bid just on keyphrases? Why not let
advertisers bid on specific positions on keyphrases?” Or, why bid at all? Just let the
advertisers pay some set amount for an ad display in a given position on the SERP.
There are several factors that prevent this straightforward type of approach. First,
the search engine does not know at any given time what a keyphrases is worth.
Second, most advertisers will prefer the top position or slot to the second slot, the
second slot to the third slot, and so on. So, there needs to be some allocation mecha-
nism, which is what an auction is. Third, as a seller of ad space, the search engine is
in the business to make a profit. So, to rank advertisers efficiently, the search engine
needs to know the advertiser's willingness to pay. The search engine uses this will-
ingness to pay as an implicit ranking of the slots.
So, having advertisers bid on keyphrases is a reasonable approach.
In addition to the bid, advertisers must also determine how much the keyphrase is
actually worth to them. This valuation will influence how much they are willing to bid.
Once there is some data, advertisers can determine this value with some calcula-
tions. Otherwise, an advertiser must use heuristics or data provided by the search
engine based on the data from other advertisers. The advertiser's valuation of the
keyphrase and the advertiser's bid are typically different, albeit generally closely
related, assuming a rational advertiser.
The valuation is the amount of return on investment (ROI) or return on advertising
(ROA) that the advertiser expects to receive based on a potential customer clicking on
the ad associated with a keyphrase and purchasing something. The advertiser must also
factor in the clicks where the potential customer does not buy anything, as these clicks
still cost. This valuation generally becomes the upper boundary for the advertiser's bid.
Auction theory [
10
,
11
] has generally shown that there is no advantage to bidding
higher than one's willingness to pay, as it is a weakly dominated strategy (i.e., it does
not get the bidder any advantage relative to any other bidder). From common sense,
this approach does not seem like a good long-term strategy either.
Therefore, we can generally assume that the advertiser's true bid is motivated by
minimizing cost, and therefore risk, by bidding the lowest possible price between the
reserve bid (i.e., the minimum bid that the search engine will allow) and the valuation
of the click (i.e., the upper bound of the advertiser's bid). So, these typical advertisers
use sponsored search to maximize the difference between total margin dollars driven
(i.e., the number of units of the product sold in a given period multiplied by the
dollar unit margin) and the advertising spend (i.e., get the most profit possible from
products sold via sponsored search), which is the base assumption of more academic
sponsored-search auction models.
However, there are certainly exceptions. For example, some advertisers may aim
to maximize order volume at a break-even point for their sponsored-search efforts.
These businesses may expect profit from repeat purchases in brick-and-mortar stores
instead of on the initial sale online.
In its actual implementation, even with simplifying assumptions, the sponsored-
search auction is a really nuanced, layered systems. Each layer is an abstraction of
the actual process. So, let us look at the sponsored-search auction progressively, layer
by layer.
