This is a guest post by PPC Hero Ally Filipe Reis. Read more about his experiments with CPC formulas and campaign settings at his Kontagiouz blog.
Recently, we noticed that we were getting a bunch of campaigns stuck in the “comfort zone.” Volume levels were struggling and there was no way to put them on the rise. You know the feeling, right? Anyway, more volume was needed and something had to change. So we wandered what to do:
“Improving qualities scores?”, “Boosting CTRs?”, “Creating new ads”, “Improving conversion rates?”… So many possibilities, where should we begin?
After some nights of reflection and lots of caffeine, we end up developing a cleaver algorithm to optimize the keywords´ Max CPC.
In a lot of posts, we all saw the famous:
Max CPC = CPA target x Conversion Rate
Ok, this seems nice! One suggestion:
If you often use this formula, please split the keywords in different campaigns by Supporters and Converters. The Supporter campaign contains the keywords that don´t convert, but assist the conversions. The Converter campaign has the ones that are mainly last click converters.
Other way, you will lose a lot of volume due to the Max CPC´s decrease of the supporters’ keywords (they have terrible conversion rates comparing to converters).
So, looking up to a Converter campaign:
Should we use the last 7 days, last 14 days or all time data? Using this aggregated data, can it mislead us or not?
At Kontagiouz, we explore Financial vertical in some European countries. Our operational level optimisations work fine taking into consideration data from the last 30 days.
Let’s check some conversion rates´ (keyword level) examples:
|Example||3 and 4 weeks ago CR (CR3)||2 weeks ago CR (CR2)||Last 7 days CR (CR1)||Last 30 days CR||Last 14 days CR|
|Keyword 1||30 %||15%||10 %||21.25 %||12.50%|
|Keyword 2||15 %||25 %||30 %||21.25 %||27.50%|
|Keyword 3||15 %||16 %||15 %||15.25 %||15.50 %|
As you can see, keyword 1 is losing performance over the last weeks while keyword 2 is gaining performance.
If we only relied on the Last 30 days, we would end up considering the same value of CR for both keywords. Not good, they are quite different!
If instead we relied on the Last 14 days, we would be giving too much weight to the recent CR!
Ok, so we developed this concept of effective Conversion Rate:
eCR = 25% * CR3 + 35% * CR2 + 40% * CR1
being, CR3 the conversion rate of 3 and 4 weeks ago, CR2 the conversion rate 2 weeks ago and CR1 the conversion rate of the Last 7 days.
Check the eCR values:
|Example||3 and 4 weeks ago CR (CR3)||2 weeks ago CR (CR2)||Last 7 days CR (CR1)||eCR||Last 30 days CR||Last 14 days CR|
|Keyword 1||30 %||15%||10 %||16.75 %||21.25 %||12.50%|
|Keyword 2||15 %||25 %||30 %||24.50 %||21.25 %||27.50%|
|Keyword 3||15 %||16 %||15 %||15.35 %||15.25 %||15.50 %|
We believe that the eCR adapts better to the evolution of keyword´s performance.
Note: The keyword must have 3 or more conversions in each period.
Very well, we have an eCR that will help us achieve the desirable CPA target, but what can we do to maximize the volume?
When we use the formula, Max CPC = Target CPA x Conversion Rate,
The value obtained for Max CPC is the value that our Average CPC can achieve for the CPA target. In fact, we always check a deviation between the Max CPC and Average CPC that is bigger with low competition´s keywords and smaller in high competition cases.
It would be awesome if the Average CPC could hit Max CPC´s value on average, so we could also maximize the volume within our CPA target.
To solve this challenge, we created a Bid Adjustment (BA) that is calculated through the ratio between the Max CPC and the Average CPC since the day after the last bid adjustment (if the bid was changed more than 7 days ago, it would be preferable to use the ratio of the last 7 days).
Note: To use the BA is necessary the keyword have more than 3 clicks in the analysis´ period.
This BA will make us increase the Max CPC so our Average CPC achieves the value we would have as Max CPC before the adjustment.
Finally, presenting our algorithm:
Max CPC = CPA target x (CR3 x 25% + CR2 * 35% + CR1 x 40%) x BA
Here goes what we got. The implementation ran from November 2013 until January 2014. During this period we didn´t made any other optimization than the Max CPC adjust at keywords level.
This campaign was created at 2012, it was very stable, making an average 679 conversions/month at a 1.57€ of CPA before the implementation of the algorithm.
November 2013 brought the change of the Destination URLs of the ads due to internal reorganisations, that’s why we had a decrease in volume.
Let´s check the values:
In January 2014, the search campaign reached similar conversion volume as October 2013, but with -37% of the costs. We also felt a good control over the CPA.
The downside is that the algorithm only takes into consideration 2 factors.
Anyway, the most important thing we try to share is to not get stuck with a fixed period of time: last 14 days, last 30 days, all time or whatever. Try different time windows to get the best adjustment of the CR trend (suggestion: use as base 7 days period, because this way you are covering all days of the week).
Regarding the BA, we are using this adjustment more and more often. If you want to increase your volume, give it a try!