Case Study: Reducing Display Network CPA by 60%
How to use bid, placement, and demographic optimization on low budget display campaigns to minimize CPA
How to use bid, placement, and demographic optimization on low budget display campaigns to minimize CPA
One of the hardest things about working with lower budget accounts is the lack of data, making it difficult to test what works - something that's basically essential to the success of a PPC campaign.
This is especially true on the Google display network. There are so many impressions available in almost any industry that your budget gets eaten up real quick if you aren't ruthless with bid management and placement selection. If you don't spend the time and money it takes to figure out what's converting, you'll end up wasting an enormous amount of money for little result.
Our budget on this campaign is just over $2,000 / month, and here's a screenshot of our CPA over time. You can see it started really high, over $190. Over time it dropped to $45.
Even on low-budget campaigns, you need to collect data. You may end up having to make educated guesses based on that data instead of reaching true statistical significance, which is fine, just make sure you have a reason for doing what you're doing.
After collecting data on this project, we noticed AccuWeather.com had just a few more conversions, a higher conversion rate, and a very reasonable CPA compared with all the other ads our sites were showing up on.
The only problem is we had tons of sites with a single conversion and AccuWeather.com only gave us 3, certainly not a significant difference.
But we did see just a few more conversions and a much higher conversion rate on AccuWeather, so at this point we had two options:
This site is in the travel industry and so it made sense that people might be checking the weather before converting, and at this rate it might take a year to get any significant data out of all these placements, so we went with option 2.
It turned out to be a good choice. We ended up adding a few other weather-related sites into the mix but that change was responsible for about half the improvement.
Next we looked at the gender report and it turns out that males were responsible for the vast majority of our conversions at a significantly lower CPA than both "Females" and "Unknown."
This was such a significant difference that we disabled "Female" and "Unknown" entirely.
Next we took a look at parental status - yep that's an option in AdWords. Turns out, parents were our cheapest conversions, unknowns were a bit more expensive but accounted for a good chunk of our volume, and non-parents were very low volume, very high cost.
Disabling "Not a parent" gave us another incremental improvement.
We could further refine our placements by reviewing the actual pages our ads were showing up on. Often, only certain pages on a site may be converting at a decent rate and the rest poorly. Cutting out those underperforming pages on our high-performing placements could help things even more.
Additionally, we could use the "common sense" approach to finding new, profitable placements should the client decide to increase the budget.
Display campaigns, out of the box, will almost always perform terribly. It's important to take the time to review all the reports, figure out which keywords, placements, topics, demographics, etc are giving you the best result, and ruthlessly trim the fat.
In the absence of enough data, start with an understanding of your target customer and a little common sense. What websites are your prospects likely to visit? What's their age? Gender? Parental status?
This will help prevent you wasting what little budget you might have to start with.