January 25, 2011
As a PPC account manager, I have the pleasure of new and exciting challenges with each new client I take on. Recently, I took on a client that was the pace setter for their PPC industry. What I mean by pace setter, is that their average ad position was 1.1 for their primary search terms. While I’ve certainly taken some keywords to top position as a strategy, I had never run into a situation where someone’s ads had consistently been running in position one and had been doing so for some time. This situation provided an opportunity to do some fairly basic PPC strategies but provided some interesting results. I thought I’d share the process with you in this case study.
This particular account is an online retailer. They sell a small number of products and the majority of their traffic is delivered on a few very specialized keywords. Conversion tracking was implemented just prior to us taking on the account so we did not have a lot of conversion-based data to jump into initially. Their key metric is Return On Advertising Spend (ROAS). For the purpose of this case study, I’m going to focus on their primary keyword, which generates more than 50% of their overall PPC traffic.
The goal of this testing was to lower their Cost-Per-Lead and their overall marketing spend. The client is less concerned about volume than ROAS so we had some room to risk losing conversions if we could continue to drive down the overall CPL and reduce marketing spend.
As I mentioned above, the few keywords that were generating nearly all of their traffic were set to bid point that all but ensured they would be in position one for their respected search queries. These keywords were all exact match, so they dominated the top position for their respected queries. While there are times when this strategy will work, I guessed that this wasn’t one of them. I knew right away that I wanted to run an experiment and play with it’s bid price. When dealing with a small sample of exact match keywords there are very few options for improvement within PPC. Your options basically exhaust at 3.
1. Write new ads to improve click metrics and conversion rate.
2. Improve quality score.
3. Lower bids to check performance at different ad positions.
In this particular instance, quality score was 7/10, not bad, and the structure surrounding the keyword was a picture of best practices. CTR was also not an issue. Rather than moving the primary keyword into a more specialized campaign/adgroup we opted to pull the non-supporting keywords out of this particular adgroup and put them into a more specialized group, which would ideally ad some focus to the existing ad group. In the end, we didn’t see quality score move.
We also wrote new ads and continue to do ad testing so this was part of our overall strategy. We’ve seen good and bad days with new ads but our top performing ad is still the original control ad.
This leaves option 3 from the above list and that is what the focus of this particular case study surrounds. Below is a snapshot of what the account looked like when we took over. I’ve cut out the keyword for client privacy but all the key metrics are there. Please keep in mind that the gray line is the adgroup as a whole while the white line is the individual keyword I’d like you to focus on.
You can see that the keyword is averaging position 1.1 and that it’s a fairly large sample size at about 10,000 impressions and 1768 clicks. From here I started a campaign experiment and reduced the bid price on this keyword by 30%. I wanted to be a little aggressive and I wasn’t exactly sure how far in front of the competition they were with their bid price. When we started the experiment we used 90/10 control/experiment settings because I didn’t want to disrupt the ecosystem of the account too much, in the event my 30% reduction was too aggressive. After about a day the experiment sample was running around position 1.9 and initial performance indicators were good so we changed the experiment sample to 70% to speed things up a bit. Once we had a sample size we felt good about, which in this case was several thousand clicks over about 3 days, we made the decision to run with the experimental sample.
Keep in mind that during the experiment itself, the 30% lower bid was running at position 1.9. I wish I had a screen shot of the experiment but unfortunately I wasn’t quick enough to realize I’d be passing this info on to you and alas, AdWords doesn’t report on past experimental samples. Here is a screen shot of that same keyword in the days following the switch to the lower CPC:
As you can see, nearly every key metric has improved. The areas I’d like you to pay the most attention to are average position and CPC. Our average position at this new bid price has dropped to position 1.6, which is actually a slight improvement from position 1.9, which is where it was running during the experiment. I didn’t put much stock into that slight increase at the time but it turned out to be the start of a trend, and the reason behind this blog post. The CPC has also dropped considerably to $2.29 from $2.90.
The final image I’d like you to take a look it is below. This again, is the same timeframe in terms of sample size and it is the exact same keyword. This screenshot, however, is about a week in the future of the second one:
Notice that in this picture, our average position has increase to 1.3 yet our average CPC has remained almost identical, dropping only 3 cents. The significance of this change is that, being the original pace setter for this keyword, we were essentially driving up our own costs. When we relaxed our bid, the competition surely saw an increase in their average ad position without much change in their overall CPC. While it’s impossible to know exactly what happened, my assumption is that our competition in turn relaxed their bids to run in about the same position that they had been but at a lower cost. It took several days on a fast paced account for these changes to take place but in the end we only lost a few tenths of a position on the average yet we’re paying 22% less for each click. Not too shabby if you ask me. The best part about this is, we’re still the pace setter, which allows room for us to run this experiment again, which we have started today. We’re not at a statistical significance just yet but I’ll be sure to follow-up with some stats when we get there.