Back before Enhanced Campaigns, mobile ad testing was very clear cut. You would simply create mobile-only campaigns, and therefore, every ad served from that campaign would be a mobile ad. This meant that ad tests could easily be performed in isolation.
Enhanced Campaigns changed this by removing the option to create device-specific campaigns. With all devices be served within a single campaign, the ability to differentiate mobile performance from desktop performance is now more complicated. Should you care? Yes, because ads perform differently based on the device on which they’re served.
As we’ve discussed, Enhanced Campaigns allow you to designate ads as “mobile preferred”. This means they’ll be given priority over your other ads to show on mobile devices (Watch out for traps with this setting). However, it’s estimated that only 3%of ads actually use the “mobile preferred” setting.
Have A Mobile Ad Testing System
One of the better strategies for ad testing — mobile or otherwise — comes from Brad Geddes –https://twitter.com/bgtheory. Brad is a widely respected AdWords thought leader and absolute master of all things pay-per-click. His method presents an organized approach to isolating and testing messaging elements at scale and advises on which metrics to pay attention to when gauging success.
1.) Create mobile-specific ad content and set ads to “mobile preferred”.
2.) Set ad rotation to “rotate for 90 days” or “rotate indefinitely” — This is critical! When given the opportunity, AdWords often makes ad rotation decisions far too quickly, which is to say, based on far too little data. Remember, Google has no direct vested interest in what users do after they click on your ads. Google gets paid per ad click. That’s partly why the default ad rotation setting is “Optimize for Clicks” (which means Optimize for CTR).
It’s nice to have the option to allow Google to make rotation decisions for you, but it’s better to make these decisions yourself based on your own unique business goals and your own determination of what constitutes statistical significance of click and conversion data. Setting ads to rotate evenly will allow you to pick the winners based on your own data thresholds.
3.) Pick ad elements to test — Test headlines, description lines, call-to-action messaging, etc. The trick here is to test elements in isolation. In other words, two completely different ads will leave you confused as to what component of the winning ad led to superior performance. This can be avoided by testing one element at a time.
As you can see, these ads are identical in every way, with the exception of the first three words of the second description line. When comparing performance of these two ads, it’s now easy to see which part of the ad made the difference.
4.) Use data over multiple structural levels — When testing ads, it’s crucial to base findings on statistically significant amounts of data to ensure you’re not making optimization decisions too quickly. One way to do this with ads is to apply tests, using the method above, across multiple ad groups. As long as all ad variations are contained in each ad group you’re including, the results will be valid.
This theoretically even works when driving ads in different ad groups to different landing pages because, by including all ad variations, you’re still giving them equal opportunity to prove themselves relative to each other. However, this can get dangerous and you’ll often want custom ads for each landing page. The important thing to avoid is pointing different ad variations to different landing pages, as this will void the results entirely.
This can be a useful technique for both small accounts in which it takes a long time to accrue data and huge accounts in which there’s simply not enough time to perform this test ad group by ad group.
5.) Set data thresholds — It’s critical to base insights on enough data. Set your own data thresholds (Brad refers to this as “Minimal Viable Data”), to ensure that you’re not going through all this work of setting up meticulous tests only to come to the wrong conclusions by taking action too quickly.
6.) Use profit-per-click (PPC) or profit-per-impression (PPI) to assess results — These are great metrics to look at for ad testing because they take all the other metrics into consideration. Here’s the problem with the others:
- Clicks — Interesting to compare if ads are rotating evenly, but easier to look at CTR, doesn’t tell you anything about conversion data
- CTR — More helpful but doesn’t tell you anything about conversions or traffic volumes
- CPC — Doesn’t give conversion data, only tells you what you paid for the click
- Conversion Rate — Even better, but doesn’t give you cost data, which includes your CPC
- Cost per Conversion — Tells you how much you paid for the conversion, but tells you nothing about conversion volume
Profit-Per-Click / Profit-Per-Impression
These metrics tell you the whole story because they factor in every phase of the user journey from all the way from the impression to the conversion. It’s a simple calculation for each:
Profit-per-click: (Revenue – cost) / clicks
Profit-per-Impression: (Revenue – cost) / impression
If you want you can do Profit-per-thousand-impressions, in which case your formula is:
Profit-per-thousand-Impressions: (Revenue – cost) / impression * 1,000
Assuming profit is the goal, this calculation will pick winning ads based on their true contribution to your bottom line, and not on any one metric that may point to success but not tell you the full story.
- Enhanced Campaigns have made it more difficult to perform mobile ad testing but it’s still very much within your power to do so effectively
- Using the mobile-preferred setting for your mobile ads is critical
- Have a system for ad testing which deals with one variable over time
- Perform tests across ad groups for more data, or to save time
- Optimize to the right metrics — PPI / PPC / PPM
- Always base findings on statistically significant amounts of data
- Remember, at the end of the day, your goal is profit, not high CTR, low CPCs, or even high conversion rates. Use PPI and PPC to monitor the true value of your ads