In May, Google began offering flexible pay-per-click bidding strategies, which allow advertisers to automate bids based upon different goals. For example, advertisers can now target a specific cost per acquisition or maximize clicks within a cost-per-click bid limit and budget.
Over the past month, my team and I have been experimenting with the target cost-per-acquisition (CPA) bidding strategy. This strategy sets your bids adjustment in real time, while taking into consideration device, location, and more. Using both historical and predictive data, Google will adjust your bids to reach your target CPA. To use this strategy, individual campaigns or ad groups must have received at least 15 conversions in the last 30 days. The Bid Strategies section can be found within the Shared Library of the AdWords interface.
Naturally, we we’re a bit skeptical about letting Google automate our bidding strategy. As a test, we set up the target CPA bidding strategy in 3 individual campaigns within a client’s account. Each campaign had a varying number of clicks; however, all met the conversion threshold. The target CPAs we set were in line with the last 38 days worth of data. For example, in Campaign A the CPA over the last 38 days was $13.28 and we set a target of $12.
Campaign C was the only campaign where we set a higher target CPA that was at least $1 higher than the previous 38 days.
After setting our target CPAs we ran a test over the next 38 days, from October 8, 2013 to November 14, 2013. Here are some things to note about this test.
- We continued to pause poor performing ad copy while writing new text ads. Flexible bidding strategies do not affect ad copy.
- We continued to add poor performing search queries as negative keywords in each campaign.
- Individual campaign budgets were not changed during the testing period.
- Campaigns A and C targeted mobile devices, while Campaign B did not.
- All campaigns targeted the entire United States.
Here are the results of our tests.
Conversions and Costs Decreased
Within all 3 campaigns, conversions dropped by at least 15 percent, and in Campaigns B and C by 41 percent. By setting a cap on the target CPA we did expect conversions to decrease, but not by the percentages in Campaigns B and C. Even with the leeway in CPA given to Campaign C, we still saw the drastic decrease. The conversion drop in Campaign A was more reasonable in accordance with the sheer volume of traffic.
Additionally, as the target CPA went so did the campaign costs. In all 3 campaigns cost decreased by at least 25 percent. In an attempt to hit the target CPAs, Google allowed cost, clicks, and impressions all to decrease significantly.
Target CPAs Were In the Ballpark
With the conversions and costs decreasing, we did see CPAs in the range of our targets. Campaign A actually did see a lower target CPA as well as a lower CPA than the previous period. Campaign B and C target CPAs were slightly higher, but only $1.66 and $1.68 greater than the previous period respectively.
The key takeaway is that Campaign A had much more data to work with to go below the target CPA. Campaign A saw 2,763 clicks during our initial 38-day time period, while Campaign B had 463 clicks and campaign C had 1,054 clicks. Since Campaign A saw so many more clicks, Google was better able to utilize the data not only to decrease CPA, but also lower costs significantly.
Both ROAS and Average Order Value Increased
Even though revenue decreased in all three accounts, in line with the conversion drop, return on ad spend (ROAS) actually increased in two of the accounts. In fact, ROAS increased by almost 40 percent in Campaign A, where we have already noted much more volume. Additionally, we saw average order value increase in all campaigns.
For advertisers looking to keep ROAS and average order value at certain thresholds, target CPA bidding may be a good option.
The positive outcomes from this test show that we were within the range of our target CPAs and two of the campaigns showed a higher ROAS. The negatives show us that we saw sizable decreases in conversions and revenue. Profitability-wise, we made good strides, but we cut out a lot of traffic that potentially could have converted.
Moving forward, we will set higher target CPAs so we don’t limit our traffic. For example, in Campaign A we might set the target CPA at $14 or $15. Even though we did actually hit the $12 target CPA, we also lost much traffic. There has to be a balance between the target CPA and the amount of traffic coming in.
We will learn from this test and continue to use the target CPA flexible bidding strategy. It did produce generally good results and is definitely a strategy worth pursuing and refining.