My last blog post took a look at calculating average position. Let’s now look at another instance where people make the dreaded “take the average of an average” mistake, calculating impression share. (See the average position post to learn more about averages of averages.)
Again, let’s take a look at some dummy data.
We have date, campaign, impressions, and impression share. The end goal is to calculate impression share by date or campaign.
This calculation is slightly different than average position. We actually have to divide, not multiply. The reason is simple algebra. We need to back into the number of impression opportunities. We have impressions, impressions share, and the formula for impression share.
| Impression Share = | Impressions |
|---|---|
| Impression Opportunities |
Multiply both sides by impression opportunities…
| Impression Share * Impression Opportunities = Impressions |
|---|
…divide both sides by impression share…
| Impression Opportunities = | Impressions |
|---|---|
| Impression Share |
…and there we have it!
Now we can apply this formula in Excel.
Once again, we can leverage the power of pivot tables and make a calculated field for true impression share. This is done by taking the sum of impressions and dividing by the sum of impression opportunities.
We can now show true impression share by date and by campaign.
There is one quirky thing I have seen in data that can throw a little wrench into this plan. Sometimes the engines lump all extreme impression share values together and only let you know that the impression share for a row of data was in a given range. For example, you might see “Less Than 10%” pop up. This is one of those scenarios that drives me nuts. I have no idea why they wouldn’t just send us the actual impression share, but they didn’t. Sometimes you just have to make the best of a bad situation. I certainly wouldn’t just exclude this data from the calculation. (It’s not as if the impression share is completely unknown.)
Without any more context into the “Less Than 10%”, we might just apply a blanket 5% for all of these records. (I’m choosing the middle here so we don’t overvalue by choosing 10% or undervalue by choosing 0%.) We can use a simple IF statement to apply this conversion in or impression opportunity calculation.
Impressions will still be our numerator in all cases, so we don’t need to make any changes there. We just need to use .05 for the denominator if the impression share was “Less Than 10%”. Otherwise, we can just use the impression share as we normally would.
Interested in working with us? See our open analytics roles here.