Reduce spend by 90%, same results
When P&G turned off $200 million of their digital ad spending, they saw NO CHANGE in business outcomes . When Chase reduced their programmatic reach from 400,000 sites showing its ads to 5,000 sites (a 99% decrease), they saw NO CHANGE in business outcomes . When Uber turned off $120 million of their digital ad spending meant to drive more app installs, they saw NO CHANGE in the rate of app installs . When big brands stopped spending on digital ads, nothing happened. Even further back in time, in 2012, eBay turned off their paid search ad spending, and saw NO CHANGE in sales coming from those sources .
Color me not surprised. When we first started developing ad campaigns for clients, running banners on a then-brand-new-YouTube.com, even when we were seeing 30% CTRs on our ads, we still found 2-5% of those resulted in site visits of less than 1 second.
Recently (nearly 15 years later), a client reported seeing the same thing: site visits that were less than 1 second.
There are only two real possible explanations for this:
- Bot traffic-based click fraud
- A website stats configuration error
Much of the problem with digital advertising today stems from marketers’ obsession with big numbers.
Outcomes matter. Outcomes are all that matters. If you are spending on reach, impressions, clicks, traffic, and visits you are wasting every single dollar, and a lot of your time to boot.
If it looks bad on the outside, it's probably worse.
To deduce where all that fraudulent traffic is coming from, and get a sense for how badly your brand can be hurt if your ads show up there, it's helpful to have a simple model for thinking about how click fraud works, from the fraudster's perspective.
15 years ago, it was fairly simple:
- Create a dummy site centered around a very specific topic (say, "All About the Benefits of Coconut Oil")
- Write a bunch of poorly-edited articles
- Submit your site to Google's crawler; get indexed
- Sign up for Google Adwords as a publisher
- Run ads on site
- Collect revenue from impressions and clicks
- A lot.
Now, let's see how this can be weaponized into massive fraud. There are several options available today:
Automate steps 1-8, above. Use OpenAI's Machine Learning text generator to create fully designed websites populated with convincing articles. For example, I seeded OpenAI with the phrase, "Some of the benefits of coconut oil include", and 5 seconds later I had the following:
And another 16 paragraphs to boot, although the narrative eventually swerved off into a discussion about chard.
No matter, this is available via an API, and creatively employed, could produce hundreds or thousands of websites populated with convincing content, per hour.
But surely no human would read these articles!
Thus, Step 2: Create an automated bot that visits these websites and clicks on the ads as a human might. This is a bit more difficult, but the techniques being used by fraudsters are more sophisticated than ever.
Also, my name isn't Shirley.
How to reduce spend and fraud by up to 99%
Automated click-fraud only works because advertisers are willing, even eager, to automate their ad spend. If you drop 10,000 paper leaflets from the sky, and 9,992 of them blow away or get washed down a storm drain, you'd likely think you'd wasted your money.
But using automated interest-based or behavioral-based targeting is exactly the same thing: online human behavior is programmatically-replicable, and by the time you've discovered it, your ad dollars have been washed down the same drain.
The solution is fairly simple, if not exactly easy:
Build your advertising on opted-in, trusted partner at a time.
Nowhere else is the phrase “you get what you pay for” more applicable than in digital marketing. If you are buying large numbers of low-cost ad impressions from programmatic channels, you’re likely getting ripped off instead of doing digital marketing.