A couple of weeks ago, Google announced a new metric in their Bid Simulator calculations, “Estimated Top Impressions”, a view of the number of times your ad should appear above the natural search results at various bids. Google may have intended this bit of data as a carrot to entice already high-spending advertisers to bump up their CPCs in order to even achieve greater prominence in the search results, but RKG has crunched the numbers and discovered an incentive to do just the opposite in some cases.
To cut to the chase, the addition of Estimated Top Impressions in Google’s Bid Simulator data has elucidated a little recognized effect of Google acting like a competitor with its advertisers within its auctions. This issue first came to light a few years ago if you could divine it from Google's carefully worded announcement of changes to its top ad placement formula, but now we can more easily assess its impact. A few months ago, RKG speculated on the high price of owning position 1 within Google’s auctions in response to a thought provoking article by Siddharth Shah at Efficient Frontier.
In addition to a few other mechanisms, we proposed the idea that Google may be “stuffing” auctions with less relevant competitor ads in order to drive up click costs for advertisers determined to own top position for key terms. Later, Shah argued that the engines may be charging a premium for top positions in order to hedge against the possibility that other ads could be more profitable for them in the same spot if given the chance. While we still cannot rule out all notions of auction stuffing entirely, we no longer believe it is a major contributor in these cases.
Google position experimentation also appears to be a negligible concern next to what we now believe is the primary agent behind the confounding numbers seen in Bid Simulator estimates: Google’s own promoted position thresholds acting like ordinary competitors. To understand what’s at play here, let’s dive into the data:
Here is Bid Simulator data for a keyword running on exact match in an average position of 1.0 for the week. These two refinements are important because they minimize a couple of variables: First, by being on exact matchtype, we know that an increase in bid will not make us available for additional broad matched ad auctions. Second, knowing that we were in position 1.0 for the entire week, we know that an increased bid will not have the effect of increasing our average position, since we already hold the maximum position for the query.
In other words, we wouldn’t be moving from position 2+ to position 1 in an appreciable number of auctions, which would cause an increase in click price by means of the traditional auction mechanics. Furthermore, this keyword is set to show on Google.com only, eliminating any impact from the potential variability of display frequency on Google’s Search Partner Network. The only variable left is the frequency with which the ad is shown in a “promoted position” above the organic listings (as opposed to being in first position but on the right rail.)
At a Max CPC bid of $2.46, this keyword is expected garner 1,780 impressions, 1,660 of which will be in promoted positions. By increasing the Max CPC bid to $18.20, the keyword should receive the same number of impressions, but is now qualified for 1,780 top impressions. Average position should not change (the keyword is already in position 1.0 at a bid below $2.46). Again, the only change is the number of times this ad shows up in promoted position.
Google estimates that the higher bid will allow for 120 additional top impressions, which are likely to bring 6 additional clicks at an incremental cost of $21 or an incremental CPC of $3.50. Besides the shock at how costly that additional layer of traffic is, one might wonder, "where does the incremental cost come from? If we’re already in position 1, we shouldn’t have to pay any more for that traffic merely because we’ve been promoted to a 'top spot', should we?"
Google’s ad auction algorithm, as explained here by its chief economist Hal Varian dictates that your CPC paid is equivalent to the Ad Rank of the advertiser directly beneath you, divided by your ad’s quality score. Since competitor data is kept constant in the Bid Simulation, we can assume that their Ad Rank remains the same and our position 1.0 term should enjoy the same cost per click whether our max CPC is $2.46 or $246 Remember, position experimentation or rotation cannot resolve the issue of our paying a higher CPC because our term is in position 1.0, not 1.1 or 1.4, etc.
Even if the reported position is a rough calculation and some experimentation in other positions is occurring at a lower CPC, a few quick calculations show that the Bid Simulator numbers cannot be duplicated by assuming our ad may show in position 2 or lower, for even a non-trivial portion of the time, and still produce the estimated overall CPCs.
Let’s take the extreme example above. Even though our reported position is 1.0, let’s assume that Google rounds down and our ad is really in position 1.099. At most, our ad could be appearing in position 2 or lower 9.9% of the time. Even if we assume we pay zero for this 9.9% of traffic, the most the CPC we pay in position 1 can be, while still hitting an overall CPC of $0.46 at a bid of $2.46, is $0.51. The $0.64 CPC we would pay at the $18.20 bid cannot be explained by position rotation. On its help page on the topic of “how much do I pay for a click,” Google outlines two special cases where your CPC is not simply: Actual CPC = (Ad Rank to beat ÷ Quality Score) + $0.01. This includes the statement:
If your ad appears in the bottom-most position above the search results, or if your ad is the only one showing above the search results, then a click will cost the amount of the CPC threshold.
The key is that there isn’t just a threshold you have to hit to be promoted, as has been made much more obvious (again with the carrot), but that this threshold effectually increases the CPC you would be paying over the traditional auction. Here's a quick example of how this could work and the huge impact on CPC it can have. Say we have an ad with a poor Quality Score of 1 for a given auction and the minimum bid threshold for promoted listing is $10.00. With our bid below the minimum ($7.00), we end up on the right rail and pay $3.55 in position 1:
Now, the CEO gives us a call and tells us to own this keyword at any cost. We bump our bid to $10.50, bringing the promoted threshold into play and bringing our ad to the center of the page above the organic results, but tripling our CPC to the full $10.00 promoted minimum:
In the absence of the threshold, our CPC would have stayed at $3.55 despite the bid increase. This can have profound effects on a PPC program particularly when you consider the notion that your Quality Score can vary significantly from auction to auction or, dare we speculate, that the minimum CPC may change as well. We’re eager to hear some feedback from our friends in the industry; there can certainly be a value attached to achieving a promoted position and the increase in traffic it will likely bring, but we also wonder how much attention is really being directed towards the incremental cost of doing so.
Thanks to Matthew Mierzejewski, VP of Client Services for his contributions to this post.