Forget Click Fraud – Stream Fraud is Even Worse!

Internet video has an ethics problem. And if we don’t fix it fast, the expected flood of ad dollars moving to our new medium will slow to a trickle – if not reverse pace entirely.

It’s called "stream fraud," and it happens when a company reports a video play to an advertiser when the video was not, in fact, viewed. Some, including Todd Sacerdoti, CEO of video ad network Brightroll, think stream fraud is twice as prevalent as click fraud — and it’s only getting worse.

According to Click Forensics, during the last quarter of 2009, more than 15% of SEM clicks were fake. That’s a big number, but stream fraud could be pushing 30% or more — and that’s just criminal.

Stream fraud takes a number of forms. It starts when a publisher claims more views than were actually consumed. This is relatively easy to do, as the accepted industry standard today for a video view is a play-start, rather than some percentage of video view completion. It results in widely inflated view counts in almost every case: According to web video analytics company Tubemogul, more than a third of video viewers bail out by 30 seconds, and by two minutes, less than a quarter of starters have stuck around.

Here at Revision3 we don’t call it a view until we’ve delivered the entire video — admittedly conservative, but I think it’s absolutely the right thing to do. I’ve written a lot about this type of overcounting of video views, but it’s only a part of the problem. At least a play-start actually shows some sort of interest, or intent, on the part of the viewer.

A bigger problem, in my book, is when video gets force-fed and autoplayed out of context and unasked for. Autoplay makes sense when it is in context and the user has an expectation that video will be played. I’m not talking about that. I’m talking about an out-of-context video autoplaying on a page, without any user initiation, and about something completely apart from the content on the page or the links that led to it.

For example, let’s say you’re searching for a new camera. You click on a link for the Canon 5D that also includes a thumbnail picture of the camera with a sideways triangle layered on top. I think it’s perfectly acceptable for that landing page to play a video review of the 5D — as long as it’s above the fold and the user can easily turn it off.

By contrast, imagine that same user clicking on a link for a 5D review, and ended up with a pre-roll autoplaying, and then a video of the 10 most photogenic nude beaches. That would be a problem — although you could make a tenuous yet plausible connection between the camera and photogenic pulchritudity.

But what if a teeth-bleaching video were autoplayed in that player, or even worse in an ad-unit in the gutter column on the right-hand side of the page? That unasked for video is now even more unwelcome, because it has no content affinity and no relevance. It’s a fraudulent video play, clear and simple.

But even worse, what if that same video autoplayed below the fold. Our camera customer would come to the page expecting to read reviews and specs on the Canon 5D, and instead an irrelevant video would start blaring through his speakers, even though the player was buried at the bottom of the page. Those 10 to 15 seconds that it would take our frustrated shopper to find — and then shut off — the offending video would be counted as a fully delivered and viewed video ad impression.

Finally, let’s take it to the worst case, and unfortunately, this is something that happens millions of times each day. What if that same video ad played below the fold, in an ad unit, with the sound turned off? It’s certainly a better experience for our shopper, as he’d be scanning specs, blissfully unaware that a sketchy dental hygiene ad was playing at the bottom of the page. But even though the video was never seen, nor heard, guess what? A legitimate ad impression just occurred. And if he stays long enough, who knows — maybe an overlay or yet another pre-roll could be served up invisibly as well.

And that, my friends, is fraud, pure and simple. It’s the act of charging someone for something that in fact was not delivered. Whatever you think about impressions, if you’re paying for a CPM, you should — at the very least — get a legitimate M.

What can you do about it? First, read Sacerdoti’s post about five ways to beat stream fraud. That’s a start, but it’s not enough. The online video industry needs to come together and support an ethics code that will stamp out stream fraud once and for all.

I had originally volunteered to put together a set of content and editorial ethics for the IAWTV, the International Association of Web Video. But now I see the problem as much, much broader. So in addition to the content side, I’m going to draft some guidelines for the distribution, advertising and placement of web video too. I encourage you to first, support the IAWTV, and second, send me your thoughts on what the key issues are here, and how they should be framed. And if you want to get a deeper understanding of the issue, and how to get what you paid for, let me know as well, I’m always happy to exchange ideas. The most important thing: let’s keep greed from killing our industry before it can truly take off!


Mark Heyert April 9, 2010

Seeing a lot of advertising partners questioning all of us because of these few bad eggs. Big fan of the TubeMogul. Not so sure about the comScore offering but…

Mandi May 6, 2010

Just wondering, do you count blog post views on this page by only the number of people who stayed on the page long enough to conceivably read the whole article?

The stats even out in presentation as long as they’re never stand alone. For example someone who lists views, ratings and comments will obviously have lower percentages of engagement than your plays, ratings and comments. If stats are delivered without context they’re never going to mean much. Also most campaigns we work on are about improving positioning so as long as apples are compared with apples there is still an integrity in the data.

NB: I’m mainly defending views here as FB video analytics give nothing so we standardise reports based on their (lack of) information and work up.

Great post, lots to think about.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.