I often get asked by new employees and potential clients, “How often do you make bid changes?” My answer is, “There never has been, and there never will be, a schedule.”
The reason? All account change frequency inherently depends on data density. First we must have sufficient volume of data on the whole, but more importantly we need to have sufficient volume at each element level. Therefore, if we are making bid changes, the frequency with which we can make data-based decisions depends on click & conversion volume at the keyword level.
Let’s compare two sample accounts of different data density. Both account spent $350K in the last 30 days across 10,000 keywords. In the first example, below, the top-spending keyword in the account during the last 30 days only spent $100.
This account is in stark contrast to the account below that spends the same about in 30 days, also with 10,000 keywords. In this case, data is highly dense for the top keywords. The highest spender is $10K.
In our data-light campaign, we are spending a lot in 30 days, but because it is all at low volume, either because of hyper-segmentation or low search volume, it’s plain to see that we have no meaningful bid changes to make. The difference between one conversion and two is so slight that we’d be purely guessing at the outcome of any bid changes, even though it’s based on 30 days.
The data-heavy account, on the other hand, could withstand multiple bid changes on these top-volume terms multiple times in the sample 30-day period, and the sample size would be large enough for us to predict with reasonable accuracy what the outcome will be. Those top three keywords account for 7.7% of the spend and 5.8% of the conversions, whereas the data-light campaign’s top three keywords only account for 0.08% of the spend and 0.07% of the conversions. So not only is the data more actionable in the data-dense account, it’s also going to be more likely to move the needle.
What does this mean for you and your account management?
1. Keep data density in mind when you are following “best practices” that preach hyper segmentation. Will you still have enough data to make meaningful decisions? If not, skip it.
2. Aggregate data across similar elements to make informed decisions. If you have thousands of ads in our A/B test with only slight variations but each ad only has a few clicks, roll them up together to determine a winner.
3. Don’t add thousands of long-tail keywords if it hurts your ability to make bid changes – be okay with some broad matching for the sake of data density.
For more information about statistical significance and a deeper consideration of the importance of meaningful data, check out this post.