One of the lesser talked about elements of header bidding is discrepancy. Header bidding discrepancies are the difference between what your own analytics report and what your demand partner reports. Simply put, header bidding discrepancies can kill your revenue by causing you to make decisions based on incorrect data. Though the header bidder is more closely tied to actual bids, which means the line items served are significantly more accurate, there are still differences in what is paid and what is served. Depending on the demand partner this discrepancy can meaningfully impact revenue.

Sources of Discrepancy

There are two main sources of discrepancy. They are ad server discrepancy and ad filtering discrepancy.

Ad Server Discrepancy

The first type is from DFP latency. In this scenario, the following happens:

  • The header bidder selects the ad
  • DFP returns the ad
  • For one reason or another, the ad is not served or not detected by the demand partner as having been served

The ad network would only record the impression at this last step. This discrepancy has many sources but the most impact comes from network issues on the client side, user behavior during the multi-step process (i.e. clicking away) and other related breakdowns along the chain.

Filtering Discrepancy

Filtering is often called “viewability” or “quality” but it is a form of filtering regardless of the name used. Depending on the ad network they may bid on inventory, then filter out that inventory later. The reasons for doing this vary, but primarily relate to technological capabilities on the demand side. Networks are not always able to pre-filter invalid inventory. They often post process their analytics. This is expected for bots, but since other types of filtering can only be done after the fact there is the opportunity for an insidious drain on the value of a network. If the network is filtering out inventory that another network might consider good inventory then giving that inventory to the more restrictive bidder is lost revenue.

Is this unique to header bidding?

Discrepancies are not unique to header bidding. The Ad Server Discrepancy type has been around forever. This is just a part of the internet. Sometimes, all of the parts in the chain don’t line up. Filtering Discrepancy has also always been around. It makes sense to post filter for fraud. Post filtering inventory based on a black box of criteria is much murkier. What is unique is the ability of an advertiser to see everything, be selective in bidding, then use post filtering to remove inventory from what they pay.

What is the fix for Header Bidding Discrepancy?

header bidding discrepancies graph
Track Discrepancies to Improve Revenue

Ultimately the fix for header bidding discrepancies is three-fold.

  1. Track your data for header bidding, DFP and your demand partners
  2. Calculate discrepancy for each demand partner
  3. Adjust your header bidding configuration

After you have a handle on the data, adjustments need to be made to auction mechanics to correct for discrepancies. We do not in any way buy into the idea that you must Recognize, Monitor and Tolerate it! You can recognize, monitor and reduce discrepancies!

How do you track & measure Header Bidding Discrepancies?

Discrepancy is ultimately a variance in reported revenue. To calculate it, you track the client side report of revenue from the header bidding analytics, track the DFP reporting for the demand partner, and then track the final accounting from the header bidding demand source. With those in hand you compare all of them to the reports from the demand source. Preferably daily, but certainly weekly. The difference between these values is the discrepancy. The resulting revenue loss is attributed to lower fill, lower CPM and lower total requests. Each metric is meaningful in terms of diagnosing the source of the issue, but the variance between each reporting source is the critical element.

How do you fix discrepancies?

Fixing header bidding discrepancies isn’t just a matter of monitoring. The percentage of discrepancy is a loss of revenue and while you may not be able to eradicate it, discrepancy can be minimized. The most direct way to manage discrepancy is to adjust the bids from the demand partner to account for the calculated discrepancy. Making updates constantly is a lot of work and making updates only every so often does not maximize your revenue. While never making updates just ends up in a regular loss.

At PubWise, we’re building functionality directly into PubWise to manage header bidding discrepancy in real time, both at the header bidding script level and through adjustments the DFP line item level. This allows for fine grain ongoing control and predictable reporting all around.