Multiple Parties In A Cookie Pool: A Race To The Bottom



The core of programmatic marketing is buying the right impression at the right time at the right price. These days, many advertisers try to find a partner that knows how to buy the right impression by giving multiple parties a chance to bid on the same cookie pool at the same time. This has several negative side effects. Higher costs for the advertiser, lack of control, poor optimization, wrong interpretation of the numbers and conversion attribution are some of the major problems that occur in such a case.

Higher costs for the advertiser. Programmatic media buying means that the price an advertiser wants to pay to show an ad to a certain user is done in real time. So, a user goes to a publisher, the bidder calculates the value of showing an ad to that user and makes a bid. The market works with second price auction, which means that the price paid to actually show the ad is the second highest price that is bid in the auction.


For example, three parties are bidding to show an ad to user A. Party X values this user the highest of the three parties. This could for instance be because the other two parties do not know the user yet or work only for advertisers for whom this user shows only little interest. Party X makes a bid of $ 2.50 and the second highest bid is $1.00. Party X wins the impression for $1.01.


In this case, everybody wins. The parties that only bid $1.00 and $0.65 do not get the impression, but they were also not willing to pay more. Party X was willing to pay more, because for Party X, this user has a high value, therefore party X wins.


Now, imagine that party X is working for advertiser Happy-fly. They run a retargeting campaign and make a bid on all users who have previously visited the website of Happy-fly and searched for a flight. The height of that bid depends on the likelihood of that user actually converting and will differ for each user. A user who has only visited the homepage is far less likely to convert than a user who has gone as far as the basket page for instance.


Now, Happy-fly wants to know if Party X is doing a good job and invites Party Z to come in and also start running a retargeting campaign on the same user pool.


Here’s what happens. The advertisers introduces a competitor for itself. Therefore, Party Z is now bidding on the same users as Party X, the second price is no longer $1.01 but has suddenly gone up to $1.76. A strong increase in price, solely introduced by the advertiser itself.



Lack of control, poor optimization and scattering of valuable data. With multiple parties bidding on the same cookie pool, there is a complete lack of control and high risk of a negative user experience. One party cannot tell how many impressions a user has seen, because they were served by another party. There is a strong likelihood that the users will be confronted with many more impressions than intended and that will harm the campaign.


When there are multiple parties bidding, however, the first party could have shown five impressions to a user before the second party steps in. However, for the second party, the user looks like a new user who has seen zero impressions from their point of view, so the second party will strongly over-value that user and place a bid which is far too high, leading to poor optimization.


The overall data collection is hindered. For any trading algorithm, it is best to have as much unbiased data input as possible in order to serve the advertiser efficiently.


Wrong interpretation of the numbers and conversion attribution.

Because the first impression served yields the highest effect, both parties benefit by bidding higher. This is a race to the bottom, because each party will bid higher and higher until the advertiser will stop cooperating with either one of them. Furthermore, users who convert are likely to have seen impressions from multiple parties, but which one gets the conversion attributed? Is that actually the party who added the most value? There is simply no way of knowing.


It is possible that one party bids the highest and wins the most conversions, simply because the first few impressions are always bought by that party. That party is not necessarily generating the most value for the advertiser, but is simply playing the attribution game best.




Work with a transparent buyer and trust them. The advertiser needs to make sure that the party that works for them is transparent about margins and media costs. This would enable the advertiser to stop worrying whether or not the buyer is doing a good job, because that can all be checked in reporting.


Take programmatic in-house. By taking programmatic in-house, the advertiser has all the tools to make the campaign run as well s possible and does not need to worry about a media buyer doing a bad job.


Split the cookie pool and divide users randomly under the different buyers. This is the most honest comparison the advertiser can make. Users are randomly distributed among the different parties. Data is still scattered and the optimization cannot be optimal, but at least the different algorithms are not influenced by the actions of others.

1 comment about "Multiple Parties In A Cookie Pool: A Race To The Bottom".
Check to receive email when comments are posted.
  1. Craig Mcdaniel from Sweepstakes Today LLC, October 28, 2014 at 12:06 p.m.

    You are likely total correct in your assumption but your ads but not be showing on a series of publisher's websites that is in the best interests of the advertiser. Also, many advertisers may not know they are themselves blocking ads to websites that would be a good match.

Next story loading loading..