A couple of months ago I read an article by Tom Hespos called "Is our ad delivery infrastructure overtaxed?" Besides doing a good job highlighting the growing issue of complexity and latency issues in our ad delivery infrastructure, it reminded me of a debate that we've had here at Adometry: is ad blocking the right approach?
A number of companies have sprung up to block ads that would appear next to objectionable content. From a brand protection point of view, we understand the appeal of ad blockers. If you could ensure that you could stop your brand appearing next to inappropriate content 100% of the time, why wouldn't you adopt one of these services? But when you look more deeply at the reality of what these services deliver and the potential unwanted side effects, the value proposition becomes less clear.
Additional Latency
Ad blockers work by inserting themselves in the ad delivery chain. Ad blockers need to make a decision whether to allow or reject an ad on a particular page without delaying unduly the delivery of the page. Most of the time, they do this by matching a URL in their cache. For a new URL, they schedule the page for examination in an "offline" queue.
How much extra latency is acceptable? While opinion varies, 100-150 milliseconds would be an upper limit, and most publishers would prefer to see something in the 40-50 milliseconds range or less.
How much latency do ad blocking vendors introduce into the ad delivery path using today's technology? Our measurement of one of the leading ad blocking vendors indicates that they add an average of almost 500 milliseconds to the ad call. Perhaps we measured them during a bad month; I’m not claiming we have enough data to be accurate about someone else's technology. I am saying that if you're thinking of adopting ad-blocking technology, you should measure for yourself the delay to the ad call.
Potential Unwanted Side Effects of Ad Blockers
Adding additional latency to the delivery of a page not only holds up content delivery, but it also means that some percentage of viewers will stop waiting for delivery and move on, reducing the size of the audience for that page. Additional latency can also increase the discrepancies between the sell-side and buy-side ad servers, something that the IAB and ad-server vendors have been working to overcome for a number of years.
Also, by inserting themselves in the ad delivery path, the ad blockers become critical to ad delivery. If the ad blocker servers are down, ads don’t get delivered.
Nested iFrames and Lack of Transparency
Another issue for ad blockers is the lack of transparency caused by multiple layers of iFrames. Publishers, networks and exchanges put ads in iFrames for a variety of reasons, some valid, some less so. Whatever the reasons for nested iFrames, they exist, and if the ad blocker can't see through them, the ad blocker can’t make an informed decision on whether to block the ad or not.
How prevalent is this problem? AdSafe Media, in their latest quarterly report, indicates that 2.6% of inventory on publishers, 17.4% of inventory on ad networks, and 31.6% of inventory on exchanges lacks transparency. These are large numbers, and as you might expect, sources with less transparency also tend to have more objectionable content. In the real world of our current ad infrastructure, it is impossible for any ad blocker to block 100% of objectionable content, and perhaps impossible to block a high percentage.
Is There a Better Answer?
It almost always makes more sense to fix a problem at the source rather than at the final distribution point. Imagine if each of us tested the quality our water in our homes or offices just before we drank it, rather than checking water quality at the water treatment plant, before it is distributed.
Some networks, exchanges and publishers have adopted tools, both commercial and proprietary, to examine their inventory at the source, before it is released for sale. This approach, which is likely to be adopted increasingly more often with the release in June of the IAB's Networks and Exchanges Quality Guidelines, can reduce dramatically the amount of ad inventory delivered next to inappropriate content, without the unwanted side effects associated with ad blocking technology.
The main downside of this approach is that it involves self-policing and can't prevent impressions next to inappropriate content for sources that don't clean up their inventory.
Conclusion
No technology can block 100% of unwanted impressions. Both approaches have their trade-offs. At Adometry, we’ve decided to partner with high quality networks to help them vet their inventory at the source, rather than take the ad blocker approach.
Good point on measuring latency for yourself. so many people simply trust what is in the powerpoint but get burned on the back end. good advice.
Both the buy and sell side have incentives to put checks in balances in place.
But some of the latency issues and the issues around new, dynamic, unseen URLs can be effectively addressed with the correct technology. (You say "Ad blockers work by inserting themselves in the ad delivery chain. Ad blockers need to make a decision whether to allow or reject an ad on a particular page without delaying unduly the delivery of the page. Most of the time, they do this by matching a URL in their cache. For a new URL, they schedule the page for examination in an "offline" queue.")
ContextWeb's RTC - Real Time Classifier - provides a real-time classification on any page URL (not depending on a sometimes dated offline "table") and can do so in 20 milliseconds. It has been "road tested" and a core part of our exchange business for years. It is also in compliance with the IAB Ad Networks & Exchanges Quality Assurance Guidelines content taxonomy.
Jay Sears
GM
ContextWeb, Inc. (ADSDAQ Exchange)
The fundamental issue within the sector has been a lack of transparency and the peace of mind that can be brought from the deployment of ad blocking technology is a necessary evil until we have an open and transparent marketplace.
Regardless of partnering with high quality publishers or not,there are still occasions where ads need to blocked. Such instances are often subjective and will vary from campaign to campaign. Take for example, the news story about a plane crash where the agency wants the campaign taken down within 20 minutes of an incident taking place. Can this be achieved manually 24/7x365, probably not ? What about the innocuous story of the baggage handlers industrial action at the airport, not big news but again, for the airline it might well be an occasion where they would want to be blocked. The point of this is that it is the client that decides where they do not want the campaigns to appear and right now, they want that extra level of protection.
As to the core issue of the article, that of latency, our SiteScreen solution is capable of providing a decision on a page within sub 10milliseconds. There is always going to be some overhead but with 5 years of brand safety experience and billions of impressions passing through our systems, we know that the cost of not blocking ads could be far, far worse.
Thanks Jay (Sears) and Ian for the comments about your products, but neither of you provided the crucial information about latency - what is the total latency that your technology adds to the page. It's not sub 20 milliseconds or 10 milliseconds. That may (although I'm somewhat skeptical) be the time to classify a page, but there are other potential sources of latency. Is the solution co-located with the ad server? If not, more latency for communication is introduced. There are other sources of latency as well. For buyers, measure the average latency of a page with and without these solutions to determine the true impact.