- Posted by neefischer
- On 2. April 2019
If you are using a bidding system like google internal bidding or an external tool you are giving up most of the control. The don’t tell you the exact algorithms – it is blackbox to the user. What i observed over time are issues like that:
- When you use Googles CPA Target there are single Clicks that are more expansive than the CPA Target?! This can not be right…
- Stacked Bid Multipliers: In a lot of accounts bid adjustments make sense – but it is getting dangerous if they are not independent from each other. I those cases you might have very high CPCs where those bid multipliers are applied.
- a low sample size is fooling the bid calculation: A high variance in basket size and conversion rate can lead to bids that are a way to high – especially when your account has a lot of longtail keywords you are probably running in that trap
So how we can check quickly if you have issues like mentioned above in your account?
First we need some data:
- The report I’m using for my analysis is the SEARCH QUERY PERFORMANCE REPORT for the last 365 days
- I added date, query, keywordid to the report so i get clicks costs and conversions on the most granular level
Without further transformation you can already see some interesting insights. Take the maximum CPC per day and visualize it over time. For a lot of customers this is the first “oh-shit”- moment when they realize that they pay a very high CPC on single clicks. The next question is normally what it means on a higher level for the account performance.
Let’s brake it further down:
We have to quantify those outlier segments based in costs vs. conversion value. A simple way to define outliers is for example:
- Take the average cpc on account level. Single clicks that are greater than 3*AvgCPC are defined as outliers
- Instead of using the mean cpc you can also have a look at the distribution. I’m using IQR-Outliers (Inter quartile range) as my second outlier definition
- Last but not least you can build a custom outlier definition with that approach:
- What is the best keyword (take the value per Click) in your account with enough sample size. Our guess is that there is no better keyword in the longtail where sample size might fool the bidding system. Calculate your highest possible click price for that keyword with your given ROI target.
Let’s put it all together: This can be your daily report for unboxing the blackbox
At sealyzer we provide a standard audit report on exactly this topic on a daily basis. If you are interested in testing our solutions just get in contact with me.