mccamon.org

The Grand AdWords Heist

“The Internet can feel like a dark, dangerous, syringe-filled alley”

I have always been skeptical of the value that online advertising brings to a company. And in particular, paid search like Google AdWords. If you haven’t given it a try it’s pretty simple: create your ads, define a budget, and determine a bid price for search terms. The results are instantaneous and site traffic is impressive, thanks to Google’s network.

So on October 21st we jumped in. SpiderOak setup a modest budget with the intention to test our way to the best search terms and then in 2016 we would expand the scope of our use of Adwords to help drive conversions of our backup product, SpiderOakONE. By mid November we turned off Google AdWords for good and completely abandoned the idea of paid search.

Why? Because of the Great AdWords Heist. Over our 21 days experiment Google reported 10,535 clicks on our ads. But our website analytics were only reporting 5,951 visits to our site. That’s right, Google was reporting nearly DOUBLE the number of clicks what we were able to verify came to our site. Wait a second. SpiderOak just spent $1,168.01 and nearly half of those clicks didn’t result in a visit? If there’s one thing I’ve learned in my 25+ years in business it’s that if something smells like bullshit, it usually is bullshit.

Google AdWords offered this explanation in their documentation; “A click is counted even if the person doesn’t reach your website, maybe because it’s temporarily unavailable. As a result, you might see a difference between the number of clicks on your ad and the number of visits to your website.” Bullshit. Our site was certainly not unavailable.

Just as we did our due diligence with AdWords testing, I dug into Google AdWords Reporting — which I can report is pretty nice — and was able to learn a little bit more about the numbers. Seems that that over 85% of our clicks were coming from the Google Display Network (not Search) and well over 50% of the clicking was happening in Romania, Brazil, Pakistan, India, Vietnam, Burma, Saudi Arabia, Kosovo, Philippines and Bangladesh. After a bit of sorting and filtering of the data I get another surprise, only 30 of our 10K+ clicks were from the US. When I compare all this info with our analytics data I see that most of the UIDs (unique identifiers of those who click) have more than one click — in fact most have dozens and several are over a hundred. Why would the same people keep clicking on the same ad?

There are good people in those countries but I have no idea why they’re clicking that often on ads that are written in English. Staring at this data I was reminded of the lively conversations about click fraud back in the early days on online advertising. Like everyone else I was confident all the issues had gotten sorted out before we moved onto more important problems like self-driving cars but then again… Bullshit.

Maybe we set AdWords up wrong. Maybe I’m reading the data incorrectly. Maybe our website was offline when 43% of people tried to visit it from far off places around the world. Maybe we were suckers who hadn’t a clue. Maybe I’m a complete idiot. Or maybe we were another victim of the AdWords Heist.

The online world can be a dark, dangerous, syringe-filled alley. Just the sort of place in the real world we might expect to get robbed. No matter off or online, safety does boil down to trust. In the offline world, we can visit the shop or have an in person meeting with a business owner to get a read on if we might choose to trust them or not. Online, it’s mostly slick marketing speak, corporate proclamations, and complicated terms of services. Mostly all bullshit.

Was it a heist? I don’t really know and I don’t really care. For SpiderOak, we now understand the value that online advertising brings to the organization, and it ain’t much. Mostly I feel like someone robbed $500 from us and we will never use AdWords ever again. We need to feel safe again online. Perhaps I should start rethinking who I trust as I navigate all these dark alleys.

Edit