This is the subhead for the blog post
3Q’s Alpha Beta process was one of the earliest innovations in SEM streamlining – consolidate your account into as few campaigns as possible based on your needs (budget, geo-targeting, CPA/ROI goals) and use 1:1 query-to-keyword mapping for greater control over your bidding.
Over the past several months, Google has updated their best practices and preached the importance of data consolidation and account streamlining over manually drilling down to top-performing segments. At 3Q we don’t like to take anything at face value, so we immediately got to work testing Google’s claims.
Here are some of the best practices and new features we tried – and how they worked for us.
New GDN Best Practices
- Use as few campaigns and ad groups as possible – don’t break out unless you have unique CPA or ROI goals for certain campaigns or have unique creative/messaging.
- Don’t create overly segmented and small audiences or divide audiences based on visit recency.
- Stick with broad audiences themed around intent/quality (Homepage viewers < product viewers < cart abandoners < converters).
- Add all ad formats and sizes (banners and responsive).
- Use Smart Bidding to allow Google to optimize for the type of users that convert best – there’s no need to over-segment when Smart Bidding will consider and automatically adjust all segments.
Did they work?
We streamlined an account that started with 50+ GDN campaigns (1 audience type per campaign) into two campaigns – one for prospecting and one for remarketing.
- We managed the old structure with manual bids. Performance was strong because we narrowed down to small, extremely qualified audiences. It was not easily scalable.
- We manage the new structure using tCPA – Google automatically optimizes to a target CPA we set and directs spend to higher-performing audiences and segments. This saves us from spending time dividing budgets by hand between multiple campaigns. We could scale spend 3x without raising our our CPCs based on how much available volume there is.
RLSA: Move from the Ad Group Level to the Campaign Level
Google introduced campaign-level audiences in December 2016. Before that, the best practice was to add ad group-level audiences to every ad group in the account. In order to see consolidated performance, you would need to download an audience report and pivot the data. Campaign-level audiences was a time saver, but most people did not immediately switch over their existing structures. Some accounts had hundreds of audiences per ad group.
This year, Google turned their focus to convincing us to use fewer, better audiences.
- Narrow down audiences to the top 5-10
- Choose a mix of RLSA, similar audiences, customer match lists, GA audiences, in-market audiences, and audiences based on visitors to your YouTube channel
- Group Audiences into Lower Funnel/most valuable (all converters, cart abandoners), Mid Funnel (creating an account, signing up for emails & engaged website visitors), and Upper Funnel (in-market audiences, home page visitors, commercial viewers, YouTube channel visitors)
- Create similar audiences off of bottom-of-the-funnel audiences (current customers, applications)
- Build audiences based on segments of Google Analytics traffic that are most engaged: long session length, high number of pages, high-performing channels
- Move audiences from Ad Group to Campaign Level
- Monitor performance and set bid adjustments
Does it work?
- Management of audiences is much easier – before, we had to download the audience report and pivot the audiences to view overall performance, but now that data is visible at the campaign level. We can make changes directly in the UI and save time.
- In one account, our audiences cover 24% of all impressions. Before we switched to Campaign level audiences, Audiences used to perform about 10% below the account average. Now Audience and overall performance are on par because we’re able to make more impactful bid adjustments.
Machine-Selected Ads and the Death of the A/B Ad Test
Creative best practices have always been based around forming a hypothesis, A/B copy testing, and setting your ads to rotate evenly so you could evaluate which ad outperformed the other manually. The process is time intensive but can produce great results if you use the correct methodology and isolate variables. If you or your client is very interested in gaining specific messaging learnings from ad testing, this is still your best option.
However, if you are more flexible with learnings or have reached a plateau with your traditional ad testing, Google’s new creative recommendations could be a great option for your account. Google uses machine learning to choose which ad to show to each user. This increases your eligibility for search auctions and drives more volume and conversions since each user is getting a personalized experience.
- Add at least three ads per ad group: Ad groups with three or more high-quality ads receive up to 15% more clicks/conversions compared to ad groups with only 1 or 2 ads
- Optimize ad rotation: choose one of the “optimized” settings to get the best, most relevant ads among users as often as possible
- Both optimization settings drive more conversions without increasing costs
Does it work?
Number of Ads
- We’ve tested this theory and found it to be true – the more ads you add per ad group, the more volume you receive. When testing 3 vs. 5 ads and 3 vs. 6 ads using campaign drafts & experiments, we found that the ad groups with more ads drove around 30% more conversions.
- Some Google documentation suggests to add up to 10 ads per ad group.
- When testing Rotate Evenly vs. Optimize for Conversions in a 50/50 campaign split on several clients, there was no contest: we saw large significant increases in CVR and conversions, as well as a smaller lift in additional volume.
- When testing Optimize for Conversions vs. Optimize for Clicks, we didn’t see a difference in performance. This may be why Google is combining the two options into “Optimize”.
As with any new feature or strategy, we recommend running your own tests to see how they perform in your account. Overall, though, initial results are promising and may eventually shift the way you think about your management approach.