January 23, 2018
Don’t Lose Sleep over Evil Bots
Earlier this month, Larry Kim wrote a post “How to Find & Eliminate 90%+ of Click Fraud in the Google Display Network”. In the very first sentence, he asserts: “Believe it or not, click fraud in PPC marketing really does exist.” And while I do agree, click fraud does exist, you will not find me huddled over my desk in a panic trying to save my accounts from evil bots. One, I believe the elimination of click fraud isn’t going to progress any account towards a goal to any measurable degree. And two, I believe in the benefits of automatic placements.
A Tangent about Stairs
A recent study revealed that more than 1 million Americans are injured in stair-related accidents each year. The math works out to 3,000 per day, or one injury every 30 seconds.
Studies also show that climbing stairs contributes to increased levels of cardiorespiratory fitness. Individuals who increase levels of cardiorespiratory health decrease the likelihood of dying early by 38 percent.
By now you are wondering where the heck I am going with this tangent. Stay with me. I’m getting there. Each time you climb a set of stairs you increase the odds that you will have a stair-related injury. But the more stairs you climb, the longer you live. So do you mitigate the risk of injury by avoiding all stairs? Is not dying as early enough of a benefit for you to climb those stairs?
My point? You don’t have to only use managed placements to have a healthy display campaign. You can use automatic placements, with the increased risk of click fraud, and achieve positive results.
All the Data
Now, before you go accusing me of drinking too much Google Kool-Aid, let’s roll through some data together and highlight some takeaways in order to spend wisely on Google’s 2 million eligible websites.
Do you know Secret Jake? I interviewed him several months ago. He’s whip-smart and as soon as he read the Wordstream post, he decided to do a little data diving (which can be similar to dumpster diving in that you never know what you will find).
Of Hanapin’s 80+ clients, he pulled any client spending more than $1000/month on the GDN. From there he pulled the stats for all automatic placements in 2017.
Here is what we found: Roughly 25% of all placements never receive more than 1 click. I am going to write it again for dramatic effect: Of the nearly 400,000 websites on which our clients’ ads were shown, about 100,000 received 0 to 1 click in the past year. Two points I want to make from this:
- There are a lot of really scummy websites on the internet where ads are showing that are not worth anything to anyone.
- While a click here or there on individual scummy sites can certainly add up and potentially eat up your budget, don’t waste your time trying to manage those one-off clicks. To Larry Kim’s point, it will take forever and be a never-ending task.
But if you take Larry Kim’s suggestion and eliminate automatic placements, you have a good chance of creating a stagnant display strategy. For one thing, if everyone were to only use what we will call “top-tier” placements, the competition is going to increase. Do you know what crowded auctions can create? Higher CPCs. But I believe the biggest downside to only using managed placements is limiting your reach, thereby causing your account to miss out on good, quality conversions.
A Closer Look
The data we will look at in this section is from a client who advertises in the publisher/sweepstakes vertical, for lead generation. When the account first started advertising on the GDN, I ran a managed placement campaign, targeting sites we knew to be relevant and have potential to perform well. I also set up a contextual keyword campaign. Neither of these campaigns had any audience targeting. In the below image, you can see results from the past year.
Both campaigns did well, with CPA coming in below our threshold of $0.40. While the managed placements campaign did perform better all-around (lower Avg. CPC, higher conversions, higher conversion rate, lower cost per conversion), if we were to have limited ourselves to only managed placements, we would have missed out on 28,297 conversions for the year. Were all those conversions top-notch leads? Probably not. But I am willing to risk some click fraud to get some of that pie.
Now, this client recognized that the value of one person filling out a lead form multiple times in a given period of time had a diminishing return. I found a targeting solution that will work to help eliminate some fraud as well.
Custom Audiences and Remarketing
I created custom audience lists in analytics and imported into AdWords. The lists were:
- Users who have converted > 3 times in past 90 days
- Users who have converted 1 to 3 times in the past 90 days
- Users bouncing from the page in 3 seconds or less
And then I layered them in remarketing campaigns. In the first campaign, I targeted any visitor who did not convert. I excluded any visitor bouncing from the page in 3 seconds or less.
I did have 10 placements I managed with bids, but I let Google to its Google-y thing. Below are the results from the past year:
Automatic placements, again, are a bit higher in CPA but still below our threshold.
I used a similar structure for the second campaign, this time targeting those who have converted at least once but not more than 3 times. I found results to be in line with the above. For me, the benefits of automatic placements far outweigh the risks.
Takeaway: create audiences using bounce rate, time on site, and other criteria to negate not only users who are not interested in your product or service but potential bots as well.
If we go back to the data Secret Jake compiled and slice and dice the data, we can find some additional ideas on how to mitigate the risk of automatic placements.
When you get into a placement list, you will see a variety of top-level domains. I do not know the statistics of the relationship of site quality with assigned TLDs. But I recommend you parse your data and look at which TLD’s perform well for you and which do not.
Here you will see that our agency-wide CPA on sites where the TLD is not a .com is 100% higher than our average CPA while our .com placements have a 10% lower CPA.
Takeaway: exclude any placement that does not contain .com (or .org, or .net; whatever your data dictates).
Next, let’s look at the content of URLs themselves. An anonymous.google placement is, to me, one of the more annoying things about the GDN. Click “see details” of the placement group and you are met with gibberish URLs to sort through. So what happens if you split the data, looking at URLs containing only text against URLs containing text and numbers?
URLs containing numbers have a 142% higher CPA than the agency average.
Takeaway: try excluding any placement containing numbers.
Listen, if you want an easy way out, then only use managed placements. But I am a firm believer you can build solid display strategy that does, in fact, include automatic placements by looking at the data. The magic is in the numbers. Test. Analyze. Repeat.