Auction-based ad serving is a clever idea that seems to have the best interests of both ad servers and advertisers in mind. Ads are ranked and costs are calculated according to simple, well-defined rules. But, as is often the case, even systems with a few simple rules can create some complicated situations.
Most advertisers on Google know that their actual CPCs are almost always going to be lower than their actual bids, and that they are paying the minimum amount for a given position. Some of the finer details get lost though, specifically in how actual CPC is calculated, which is:
(Next lowest Ad Rank / Your QS) +$0.01
The key feature of this is that any bid adjustments that do not move you above or below an adjacent Ad Rank threshold will not affect your CPCs. This is often a relief to advertisers aggressively bidding in the top position, as they can rest assured that they are only paying the minimum necessary.
The problem with this is that it can leave you open to increased CPCs due to competitor activity. To illustrate this, let’s take a look at the following simplified scenario for 3 advertisers bidding on an exact match term:
This is characteristic of branded auctions where Advertiser A is aggressively bidding on their own brand terms and Advertiser B and C are product resellers. Everything looks pretty standard; the brand in question is firmly cemented in position 1 with a much higher Ad Rank and a CPC less than the competition.
However, there’s an issue here: Advertiser A is overbidding. As discussed, the way Actual CPCs are calculated means that you will never inflate your own CPCs at a given rank by increasing bids – but that doesn’t mean that competitors can’t. Check out what happens when Advertiser B raises bids:
The ads are still being ranked in the same order and Advertiser B’s CPC is unchanged, as it is set by the competitor below. The only change in this auction is that Advertiser A’s CPCs are greatly inflated.
It’s worth noting that as ad ranks become closer together, Google seems to sometimes change up the order of ads served, so it’s likely that Advertiser B will receive a small amount of impressions coming in at position 1 (with clicks costing $2).
Advertiser A might be more than happy to continue to maximize volume and pay $1.41 per click, but there’s still no reason to overbid. If Advertiser A’s bid was closer to the actual CPC to begin with, increased activity from Advertiser B would result in an increase in Advertiser B’s actual CPCs, Advertiser A would get more data on CPC vs. CTR by rank, and a standard bid war would develop. The result might actually be the same, but Advertiser B would need to be with paying their increased bids at every step. In extremely competitive auctions with only a few key players, you can often run into huge drops in CPC by position; in these cases top-position advertisers might consider lowering bids to ‘keep competitors honest’ and force them to pay their bids.
To summarize, aggressive overbidding is nearly always a bad thing, even if you are willing to pay CPCs approaching much higher bids. By keeping your bids near your CPCs, you dis-incentivize competitors from inflating auction prices – which can even happen unintentionally. AdWords campaign experiments can be an incredibly useful tool for determining CTR, CPC, and CPA at different places in the auction, effectively giving you data points on both profitability and scale.