Sitting in dingy poorly lit rooms in covert office blocks in Estonia, Cairo, Dhaka and New Dehli, workers work through the night, doing a menial task for a measly wage. But rather than stitching together high-end sneakers or sewing buttons onto $200 pairs of jeans, the workers in this case sit in front of glowing computer screens, clicking on the paid advertisements that appear on websites. These click farms are ad fraud—a problem that has plagued online publishers since advertising first migrated online. And, as technology advances, fraudulent methodologies are becoming more sophisticated, turning this problem into a large-scale black market with potential to seriously undermine the value of the online metrics digital evangelists so often champion.
In more modern iterations of these fraudulent setups, the scammer might create a website and then start feeding ads onto it through the Google AdSense programme. Next, the scammer introduces a bot that emulates millions of human clicks per day, resulting in the fraudulent user account being credited.
Lindsay Mouat, the chief executive of the Association of New Zealand Advertisers (ANZA), says that research into this issue has already painted a troubling picture for the industry.
“A survey by ANZA’s US partner the Association of National Advertisers in late 2014 suggested almost a quarter of video ad impressions and more than half of third partysourced traffic is fraudulent,” says the ANZA head. “The study by online fraud detection firm White Ops analysed 181 campaigns from 36 major brand advertisers, which were tagged to identify bot fraud. During the study, 5.5 billion impressions in three million domains were measured over 60 days in line with industry spending patterns.”
The topline findings of this study reveal that large chunks of the impressions delivered were in fact facilitated through fraudulent activities: “Botnet controllers hijack everyday consumers’ identities and home machines to conduct ad fraud; almost a quarter (23 percent) of video ad impressions were identified as bot fraud; 11 percent of display ad impressions were classified as bot fraud; publishers who bought sourced traffic from a third party as a means to drive additional unique visitors to their site, had a bot fraud rate of 52 percent on that sourced traffic; programmatic display bot traffic averaged 17 percent bot fraud; and bot fraud for retargeted ads was 19 percent.”
What is most troubling about these stats is that advertising is increasingly being sold—especially in programmatic—in accordance with the number of impressions (or clicks) that are served. And if a large percentage of those clicks are actually generated through fraudulent means, then this in turn means that advertisers are paying for more than what they’re actually getting.
Given the money involved, this is something that the industry can’t ignore. The recent figures from the Interactive Advertising Bureau (IABNZ) showed that advertisers spent $589 million on digital advertising over the course of 2014, leading the body’s chief executive Adrian Pickstock to suggest that interactive advertising had “almost certainly” overtaken newspaper ad spend. So even if the percentages quoted during the study are on the heavier side, this could still mean that a significant number of ad dollars are being paid toward bot rather than human views.
Can video ad fraud derail programmatic?
According to TubeMogul’s Q4 results, New Zealand’s programmatic video advertising market grew nearly three-fold the fourth quarter of 2014, with video ads available for real-time buying jumping 210 percent to 260 million auctions, from 90 million in the third quarter.
“Automated buying in New Zealand is advancing rapidly, underpinned by the desire of brand advertisers to drive transparency and efficiency in their ad spend,” said the company’s Australia and New Zealand managing director Sam Smith in a release at the time.
The problem, however, is that programmatic is often priced in terms of cost per thousand views (CPM). And this obviously isn’t ideal when statistics suggest that as much as 25 percent of all online video views are registered fraudulently through bot activity.
As illustrated by the growing demand in the programmatic market, this hasn’t yet stopped advertisers from spending ad dollars on programmatic ad-buying.
As OMD’s Andrew Reinholds points out: “Ultimately programmatic for OMD delivers such a cost-effective audience for our clients that even say if 25 percent of the clicks were fraudulent, the way we track and optimise, then the remaining 75 percent of the actual clicks are of real value for us and the CPA [cost per action]delivered will be well below other types of buys.”
And provided that programmatic players can limit the extent of ad fraud and keep this perception of adding value to the client’s offering, then it’s likely that the market will continue to grow.
A measurement problem
In his summary of the ad fraud study, Mouat urged the industry players to start working on a strategy to enable accurate measurement of online impressions.
“The industry needs to collaborate—advertisers, agencies, publishers—to substantially reduce bot fraud,” he says. “This [demands]transparency for sourced traffic, inclusion of language on non-human traffic in terms and conditions, being transparent on anti-fraud policy to all external partners, the use of independent monitoring and an ongoing analysis of fraud in advertising traffic.”
In traditional media channels, tracking of campaigns has relied on industry research conducted by independent third parties that aim to determine the number of readers, listeners or viewers that engage with a medium.
In the case of television, Nielsen provides data on the performance of shows; in radio, TNS has until recently conducted a twice-yearly survey on the number of listeners that tune in; and the magazine industry continues to rely on Nielsen and ABC figures for readership and circulation, respectively.
In each of these instances, media owners sell their ad slots based on information that’s provided by the independent third party. However, as things stand at the moment, the digital industry does not have a preferred supplier for online ratings, making it challenging for advertisers to determine what the most accurate data is.
To address this problem, the IABNZ last August released a document that compares the different measurement vendors available in New Zealand.
“No current audience measurement system is perfect – each has its own strengths and weaknesses so the Measurement Council helps progress improvements across multi channel platforms and keep New Zealand at the forefront of best practice,” says the online publication.
Nielsen, Comscore, Experian and Effective Measure are all pitted against each other in terms of methodology, metrics, positives and user interface, among other things. The IABNZ is not hedging its bets on any specific vendor, and instead says that its “approach is to provide the industry with objective information on the existing methodologies.”
In doing so, it places the onus on advertisers, media agencies and publishers to incorporate their own measures to ensure that their reporting is accurate.
A media agency yoke
Given media agencies’ role in buying the most suitable slots for their clients, it follows that they are principally responsible for safeguarding against fraudulent activities that may obscure the truth behind a campaign’s performance. For this reason, media agencies across New Zealand have consolidated their online campaign tracking capabilities and skills.
“Bot traffic impacts every single campaign online,” admits Nick Boulstridge, the head of digital media at VivaKi. And he says his company has responded though VivaKi Verified, a policy that sees VivaKi evaluate and score web advertising platforms in terms of how they monitor and filter out bot traffic and other fraud, as well as their process for vetting sites for brand safety and quality executions.
“We’ve done a lot of testing on this and found that our whitelist within AoD, our programmatic platform, does a very good job of keeping us off this sort of inventory,” he says.
Similarly, OMD’s managing director Andrew Reinholds says that his agency has also introduced a strategy that helps to identify fraudulent activity.
“When using programmatic buying we utilise cost-per-click (CPC) deals the majority of the time, which therefore sets us up to a risk of being charged for clicks that a ‘non-human’ has generated,” says Reinholds. “The way we counter this is by tracking from click-through to session generated on the website and then through to the desired action for the campaign.”
Taking these systematic steps allows OMD to spot anomalous trends—such as an uncharacteristically high bounce rate—that are consistent with bot activity.
Although the approach is proving effective, Reinholds says that the fight against ad fraud isn’t easy due to the levels sophistication involved.
“It is always going to be difficult to categorically identify the volume of ad fraud,” he says. “It’s been going on for years with search and Google rightly, when they identify suspected fraudulent activity, reimburse their clients. That said, there could be so much more going on that we don’t know about.”
The greyest area
One area that is proving particularly difficult to track is video advertising and Boulstridge believes that this is largely because of the system currently employed.
“Video reporting lags behind due to limitations within the VAST standard,” he says.
Last year, in an article published on AdExchanger.com writer Ryan Joe explained that if a “video was integrated using only the video ad-serving template, then marketers might as well dump their video ads into a titanium box for all the visibility they’ll get.” He went on to say that VAST video placements “don’t allow third-party tech vendors or advertisers to apply viewability technologies,” making it difficult to tell if an ad was viewed by a person.
One way to get around this problem is by incorporating a video player ad-serving interface definition (VPAID), which would support deeper analytics of the video ads running via a platform.
“Unfortunately in New Zealand, VPAID inventory is quite limited due to the dominance of YouTube (Google has rejected the VPAID standard),” says Boulstridge. “True viewability and verification is only really possible with VPAID.”
A Google spokesperson disagrees with this statement: “Google’s ad exchange allows buyers to serve VPAID ads to publishers that opt in.”
And while this might be true, the quick look at TrueView policies for ads served onto it explicitly states that “VPAID is not allowed on YouTube.” And a spokesperson from programmatic ad-buying platform TubeMogul says: “Except in certain circumstances, YouTube generally does not accept the VPAID format, which limits third-party reporting. Generally, [TubeMogul] supports publishers and supply-side platforms transitioning from VAST to the VPAID standard as it allows for more robust measurability.”
What this means is that advertisers are more often than not reliant on the statistics provided through Google’s interface and do not have recourse to independent verification of statistics.
“Advertisers have full transparency on the number and duration of views on YouTube, as well as other metrics,” says the Google spokesperson. “And using TrueView, advertisers on YouTube only have to pay when a viewer watches their video for 30 seconds, or to the end if the video is shorter.”
But Boulstridge isn’t convinced by this argument. “It is very easy to say that TrueView has delivered and only charged for the completed view, but was it human? Without third party tracking you have to accept what is provided by Google.”
Time for standardisation?
“Having a media owner responsible for the tracking and measurement of campaign performance is unusual,” says Reinholds. “By default it can create a perception of non-impartiality and conflict of interest. As with all other media, an independent third party means that these perceptions are removed and everyone is working on a level playing field.”
This is not to say that new media juggernauts such as Google and Facebook aren’t fighting the good fight against fraudulent online activities. Both organisations have invested heavily in technology that identifies ad fraud, and it is, after all, in their favour to have platforms that advertisers trust enough to spend money on. However, there is something icky about giving the publisher exclusive right to report on campaign performance.
So does this mean that it’s time to standardise the industry by incorporating third-party tracking facilitated by an independent third party?
Sam Smith, the Australia and New Zealand managing director of TubeMogul, believes that standardisation does have its merits but could also inadvertently slow down the battle against ad fraud.
“There are situations where standardisation by industry bodies is helpful,” says Smith. “The standards adopted for video viewability adopted around the world obviously helped advertisers have a single metric for whether a viewer had an opportunity to see an ad and what that means. That said, there is a wealth of data available to advertisers in digital advertising and we should not detract from the diversity of what is available by being too rigid as an industry.”
The very diversity of approaches currently being used by industry players means that fraudsters are fighting on myriad fronts—and this makes it more difficult for them to game the system.
However, there is also room for collaboration in this fragmented approach. As Smith explains: “We believe the best approach is to develop your own technology and also work with other leading players in the market.”
But even if the industry does collaborate and find ways to effectively mitigate the problem, there’s no guarantee that ad fraud will disappear. In much the same way that sweatshops are still producing fake products, there’ll always be unscrupulous characters who find ways to make illicit profits online.
- This story was originally published in the March/April edition of NZ Marketing.