The Google Display Network can be a really messy place, especially if you’ve got multiple campaigns and targeting options bidding on automatic placements. Odds are you’ll end up showing ads from multiple campaigns/targeting types on the same sites across the GDN. Although this might not seem like a massive problem, there’s most likely a best targeting option for each site. One targeting type that out performs the others that you should take advantage of. But how can you determine what it is? That’s what I’m hoping to help with today. At the end of this post you should be able to evaluate performance of automatic placements triggered by multiple campaigns and determine which targeting method has the best performance.
Export Account Level Autoplacements Report
Easy peasy. Head on over to the Display Network tab and choose Placements. Remove the ad group column from the report. If you don’t you’ll be given a new row for each ad group the placement has shown for and that can cause a problem when looking for duplicate listings later. Now download the report for the date range you would like to analyze. Depending on traffic, I usually choose 3 to 6 months worth of data but you can be flexible here. For accounts with low conversion volume you may need a bit more and higher volume accounts might need a little less.
Highlight Duplicate Placements & Delete the Rest
Since we’re only looking to determine optimal targeting for certain sites, all sites not being shown for multiple targeting types are irrelevant. To make the data easier to work with, we’ll need to remove all sites not being targeted by multiple types. Luckily Excel comes with a handy little preset filter to help out. Highlight the Placement column, then highlight duplicate values using Conditional Formatting.
Once you’ve done this step, all of the sites that appear more than once in your report will be a new color of your choosing. Filter your placements by color and remove all of those with no fill color, leaving only your duplicate sites listed.
You can double check if you’ve done this right, because if you’ve deleted all of the cells without color sorting by “No Fill” will be gone.
Now your’e ready to start analyzing the data. Wahoo!
Create a Pivot Table & Analyze
In the pivot table, you’ll be able to see all of the aggregate level data for each one of your placements. Since the end goal is to determine which targeting type works best for each placement, set up your pivot table according to the screenshot below:
This will give you a view something like this:
From this view, you’re able to see which targeting types work best for your ads on each autoplacement. You can now calculate CPA, CPC, Conversion Rate, etc. and use your own prioritization to determine which type of targeting produces to optimal performance for your campaigns. For example, some might view the performance of Retargeting for youtube.com as the best due to the highest volume, but others might see Similar Users as a better option due to good conversion volume and a significantly lower CPA.
One thing to remember when looking to compare data, do NOT rely on CTR, CPC, Conversion Rate, or CPA calculations within the pivot table. That will only add or find the average of those stats from the raw data. It will not weight those stats based on volume. Instead, you should calculate those stats by typing the cell fields from the pivot table into the columns to the right of the table. For example, if I were to calculate CPA of each line it would look something like this:
And voila! Consider your autoplacement duplicates untangled! The actions you take on this data are completely reliant on your account goals. Be sure to weigh the options of volume vs efficiency when deciding how to optimizer your account.
Fun fact: this process can also be applied to determining which ad group (targeting subtype, aka set of keywords, topic, interest category, etc) performs best for each site as well. Simply download a campaign level report and add Ad Group to your Row Labels instead of campaign. See, that fact was pretty fun, eh?