Your Ad Server, My Ad Server: Trends With Discrepancies Today

DiscrepanciesFor many years, the deliver of display media impressions through ad serving on a single media buy had resulted in a disconnect between what two different parties said was delivered in the ad serving food chain – think advertisers/agencies, ad networks and publishers .. they’re are all in this mix.  Industry initiatives and innovation has continued to drive at solving the discrepancy issue. So, where are we today?

We asked several industry executives to chime in with their views on the following question:

“What’s your take on trends you’re seeing with discrepancies in ad delivery reporting today?”

Click below or scroll down for more:

Mitchell Weinstein, SVP, Director of Ad Operations at Universal McCann

“Discrepancies have not been a major issue for the past couple of years at our agency, ever since the revised IAB/AAAA T&C document was released. As long as we are using an industry accredited ad server, we have been able to use our agency ad server counts for billing. That’s not to say discrepancies don’t exist between agency and publisher counts, because they do. But the back and forth over how to handle billing has been greatly reduced.

One area where discrepancies still cause some issues is between the different agency ad servers. We will often use multiple ad servers for a single campaign (standard, dynamic, rich media, video, etc.), and those ad servers will never agree 100% on the impression counts. So it’s important to identify up front where the billing numbers will come from. Complete and open communication with our media partners is key to making this work properly.”

Jay Wright, Yield Management Group Leader,

“For, the trend has gotten better. But that’s only because we switched to DFP and most of our advertisers use DFA. Interestingly, this fact brings up a whole different issue. I find that the advertiser’s choice of agency side server makes a huge difference. For a while, we were tracking discrepancies by 3rd party system. Probably not surprisingly, different 3rd party servers have discrepancies with completely independent (and predictable) means and standard deviations. Some are normally distributed around 3%, but others are predictably 5% or 7%.

I’m completely in favor of accountability and I’m more than willing to have my site be held responsible if we don’t deliver on a contract. However, I’m not willing to lose 5% extra revenue because of the agency’s choice of ad server. The short term fix is to simply pad the impression goal up by the estimated loss from a particularly wasteful server. But often, those padded 5% or 7% of impressions are in valuable content areas that could be monetized by other campaigns. I understand there are different business reasons for agencies to choose different 3rd party systems, but it seems very unfair to push the cost of that decision onto the publisher.

In the long run, there are only two ways I see to be fair to the publisher (assuming the publisher is reputable and they are using reputable technology too). The first is to pay on first-party numbers. The second is to have a rate card based on the agencies choice of ad server.

For the time being, this doesn’t seem pervasive enough to take either of those actions. But, if we continue to see agencies running increasingly elaborate creative on new platforms and technology, the variable pricing is something that I’d strongly consider implementing at our company.”

Bart Boughton, VP, Advertising Operations,

“Unfortunately, discrepancies between publishers and 3rd party delivery systems continue to be a problem despite advances in reporting and various efforts to bring greater transparency to the numbers. This becomes magnified as publishers’ inventory mix extends beyond standard banner units to rich media, video, mobile, and tablet inventory. Currently, publishers must manually aggregate reports to compare primary (internal) delivery numbers with 3rd party delivery. This can be a tedious process that is difficult to maintain and publishers with small teams often find themselves discovering large discrepancies and other issues well into a campaign rather than at launch.

One way we combat these issues at TheStreet is by creating an identifier for each ad/placement in our primary system that gets tied to a placement or placements in the 3rd party system. The 2 are linked in each report shortly after a campaign launches, then going forward the comparison can be analyzed on a regular basis. There is an effort required to create and maintain the initial association, but the benefits far outweigh the investment of time. Publishers can detect and address discrepancy issues early on, view placements that are not live that should be, and perform accurate revenue forecasting using billable numbers rather than internal estimates.”

Mike Lewis, CEO, Ad-Juster

“Discrepancies have not and are not going away. While display discrepancy rates have come down some in the past few years, the reality is there is still a reasonably large average discrepancy that varies by adserver and depends on an array of other factors including creative size, 3rd party ad vendor, number of total files, and number of trackers and cookies. A common strategy to compensate for these discrepancies is to use complex rules based on loosely correlated factors to set local campaign buffers. Unfortunately, the significant variance from ad placement to ad placement often defeats these static approaches. This failure leads to either over or under delivery which eats directly into publishing margins. Our initial analysis with several publishers has produced some eye opening numbers about exactly how many impressions are effectively “thrown away” each month using similar schemes, and exactly how much potential revenue they represent. The only real solution remains active, preferably automated, buffer management by the ad operations team. Video, rich media, and ultimately mobile increase this imperative due to their significantly larger average discrepancy rates.”

Daniel Davies, Director, US Ad Operations, Adnetik

“The disparity between reporting touch points (e.g., ad server, exchange, DSP) seems to be showing some improvement over the last year, after some very discrepant times.  However, there still exists a fundamental difference in who’s counting an impression when.  Few veterans of the digital industry would deny that things have gotten much more complex over the last five years, if measured only by the  amount of activity occurring within that split second before an ad is served.  Data providers, ad servers, DSPs, exchanges, aggregators, resellers, each recording and interpreting the same event a little differently.  What we sometimes end up with is a handful of varying snapshots, all taken of the same thing, but at slightly different times.  And,  it seems as if the number of entities involved in a single impression rendering is only getting larger with each year.  So, one would assume that the discrepancy in reporting would, likewise, be growing.  But the computing muscle and stamina backing digital advertising is bulking up, as well.  Larger tech stacks, lower latency, and greater overall industry know-how are all helping to keep the tremendous potential for discrepancy somewhat at bay.  This is a trend I think we all hope persists into 2012.”

Manu Warikoo, SVP of Solutions, Operative

“Discrepancies in ad delivery reporting are a persistent problem that impacts all publishers in the United States today. Publishers are spending a lot of time and money trying to reconcile the data and are looking for easier, simpler ways to address the problem.

We see several trends that indicate this is still a major challenge for publishers. For example, while the percentage of total line items billed as third party is roughly the same from 2010 to 2011, the percentage of contracted spend that this represents has significantly increased YOY. What this suggests is that larger agencies are influencing publishers to bill on third party ad servers.

Publishers of all sizes are affected by these billing terms with many billing well more than half of their revenue against third party numbers. Some publishers however, appear to have a reduced impact from the issue by billing exclusively on primary delivery. So while it seems that there some holdouts, we do see more aggressive movement in the US market towards billing off of third party numbers. This is less of an issue in the UK because most publishers there are billing on primary ad server data.

Many publishers are automatically padding their impression goals on their primary ad server to account for the discrepancies. This can be anywhere from 5%-20% depending on the third party ad server.

To help alleviate some of the burden and additional overhead/resources required to reconcile third party discrepancies, we are seeing a lot of demand for automated solutions, especially ones that integrate with an order management system. The whole reconciliation process requires a tremendous amount of manual time and effort, which can impact a publisher’s profit margins and time-to-market for new initiatives.”

By John Ebbert

More AdExchanger “Industry Reactions”…

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Interestingly, no one mentioned the lack of API access to near-real-time reporting from the agency ad servers. While i understand there may be valid business reasons for not making these generally available, it is a disservice to both advertisers and publishers to not provide what most platform providers would consider a ‘basic’ API service.

    Currently, most automated solutions involve html parsers (page scrapers) or e-mail ingestion. While these are not complicated to build, they are very limiting for automation purposes (stability, time delays, etc). In the end, this means more resources are required to do basic functions like deliver, report and invoice a campaign. Which ultimately means higher costs to all parties involved.

    In the end, until the agencies push for better features from their platform providers, it seems we are stuck with the status quo. I like Jay Wright’s suggestion that rates cards be adjusted by agency ad server. Maybe this could be the catalyst to better integrations.

  2. So long as ad servers embed redirects as ad calls, there will always be some discrepancies. Sadly, this is an inefficient process the entire industry is based. With every redirect there is a loss of impressions from latency and other issues.

    The rule of thumb I’ve seen most common is to accept 10% as “ok”. Anything more needs to be reconciled, but as some point out, this is not very common.

    A bigger question is to revisit the need for advertiser’s to use a third party ad server or using multiple ad servers. The basic DFA and Atlas third party servers have not evolved much since they were created in ’90s (yes, about 15 years ago – eons in the online world). As a result, specialty servers, measurement tools and trackers have sprouted up. Now ever ad call includes all these.

    Tagging and ad server operations are grossly inefficient and need to be streamlined with some tech.

    Think about it – shouldn’t there really be one solution, one tag, that can accomplish all or most of these for an advertiser? Why do I need to install a third party tag to verify where my ad ran – isn’t that a core value of a third party ad server – verification? API is a great start – if you integrate servers you don’t need more tags.

  3. It seems like no one is talking about the elephant in the room. Discrepancies between DFA/P and most ad servers (MediaMind, PointRoll) is not a big deal. Atlas always seems to be the one with huge discrep numbers. If those few agencies still using Atlas moved to another alternative I don’t think we’d even need to have this conversation.

  4. Discrepancies occur at every level, but it takes a certain amount of intelligence from third party ad servers to understand what a duplication of events looks like so it can properly de-dupe it. This requires an incredibly robust infrastructure that reduces latency across sites, tags, the server itself and the data management platform.

    To say 10% is “ok” is bad if you spend a lot. 10% of a million dollars in ads is $100k. It’s worth it to invest in a quality third-party ad server that actually can reconcile duplicate events on the fly rather than rely on spreadsheet kung fu to get it done.

  5. I’ve been railing against discrepancies and writing about how to manage them for years now – it really is the infrastructure skeleton in the closet for this industry, and typically dealt with in a really lazy manner.

    The fact of the matter is, the publisher is the one that gets stuck holding the bill, no matter what the IAB T&Cs say, so not many ad servers have an incentive to do much about it – we’ve all just come to accept them as a cost of doing business. Change is likely not in the works unless as Jay suggests, publishers price their inventory based on an advertiser’s 3rd party and impose a cost on their clients, which I’d love to see, but doubt is all that realistic.

    The practical reality is discrepancies are a management problem, how do you efficiently and reliably identify them in advance, and manage against their impact to revenue, while minimizing the opportunity cost they have on your inventory. It’s a problem worth looking at if you’re a large site – if you look at how much inventory you need to clear your discrepancies vs. what you actually use to clear them and add that to your lost revenue for unexpectedly large issues, I bet it would shock most people in the sales organization.

    As a starting point, people should look at Mike’s company if they aren’t already clients – it’s a pretty great solution toward getting a handle on the discrep problem, measuring them, and putting a process around dealing with them.

  6. James Perlskogg

    I see that everyone agrees that the pub side could use some help as far as leveling the information field – so to speak.

    As of now, a large set of tools are developed from the point of interest of the marketer/advertiser…look at something as a simple analytic metric: unique impression..that tells a marketer that one impression is served to a different audience member than another unique impression..which allows them to work across publishers to gain reach by capping impressions to a specific/unique audience member….increasing value.

    ..the PUB side needs a unique publisher metric, that shows which portion of their audience doesn’t visit any other site and is able to price that impression accordingly. Is anyone aware of a company that does that outside of the basic unique audience data provided by ComScore?

    Stop trying to make serving the ad more “efficient” and “centralized” and help us to understand what the space is worth in the first place. And by understand, I mean give us information and we’ll give you a price, don’t force us into a open exchange and pit us against each other.

    If this keeps up, I see a “publisher union” forming rather quickly.