It’s been a few weeks since my last blog, Clicks, The False Positive. Since then, more musings have surfaced in the media about the fiction associated with the click. My favourite was a recent article in Business Insider by Shann Biglione of Zenith Optimedia. He went after the click with more vigour than I’ve seen before. Some of his words of wisdom: “We have produced a click-obsessed monster so ugly that even the people who work in advertising cannot stand its stench.”
To Shann’s point, many things stink about the click. Last time I briefly touched on how ad fraud degrades the click even more. It’s a serious topic with serious revenue tied to it. As marketers shift more budget to digital media and adopt programmatic solutions, and as channels of consumption continue to blur, fraud acts as an opportunist—computer-optimised to engage when the opportunity is ripe—and the losses can multiply quickly.
I’ve sifted through oodles of research on fraud. While the numbers vary, a recent study by the Internet Advertising Bureau (IAB) estimates the total annual cost of flaws in the internet supply chain to be $8.2 billion. Invalid traffic tops the list at 56%, or $4.6 billion in losses.
Fraud costs
Click fraud sits in the invalid traffic segment. And it’s not just the financial waste that’s the issue: It’s the downstream impact of having bots in the analysis pool of clickers, which is already small. Organisations that emphasise click performance for display engage in a variety of behaviours whose consequences erode understanding of the marketing mix. Here are a few I’ve either heard about or observed first hand:
| BEHAVIOUR | CONSEQUENCE | |
| Presenting display performace based on clicks to justify program spend | > | Misses the vast majority of display impact, over credits other marketing program, favours carpet bomb method to drive clicks regardless of relevance/value |
| Including display program clicks versus site visits in click based attribution schemes | > | Fails to connect view through to conversions, undercuts the value of the display impression, drives flawed decision making on display budget allocation |
| Modelling or optimising clicks | > | Using good science with bad inputs - not only a small sample but a flawed one because click data includes fraud —bots are more predictable than people |
| Not requiring SLA's with agencies, publishers or vendors to vet click fraud | > | Wasted budget, revenue loss, and deviations in analysis related to clicks |
| Overlooking the Unique count of people when evaluating based on clicks | > | Basing success on a narrow group of individuals and not evaluating or understanding consumers in the view through group. |
You may be wondering how, with all the advancements in technology, companies work to mitigate fraud. Some of the more talked-about include viewability and other guarantees.
But the key strategy is understanding what you may be bidding on. And that’s done by learning from a strong, robust network that tracks people over time, like Conversant does. The network sends signals that help decide who may be the most viable to deliver an impression to. Some of the red flags we look for:
- Dead cookies
- People who never transact
- People whose only activity is clicking
- People who are only active between 10 and midnight every night
(indicating an artificial surfing pattern) - Any other patterned clicking activity that appears to be automated
These are just some of the signals of bot behaviour. A network that can recognise them can mitigate fraud and preserve advertiser budgets.
So, even more reason to get off the click bandwagon. And next time, instead of talking about the worst way to measure display, I’ll talk about the best way: accounting for the view-through and measuring incremental sales.