Join us April 28-30 | Austin, TX

Learn More

Profit per Impression: A Better Metric for Ad Testing

Print This Post Print This Post

Today we have a guest post from one of our PPC Hero allies! Michael Wiegand is a PPC Guru & Google Analytics Consultant at Portent Interactive, a full-service internet marketing company based in Seattle, WA. In this post, Michael explains why Profit per Impression is the perfect metric for ad testing.

—————-

Ad split-testing is one of the core disciplines of PPC. If we don’t test, we can’t improve. And as paid search professionals, we become obsessed with finding the right combination of headline, call-to-action and destination URL to influence better account performance.

But there’s a problem with this discipline, especially from an e-commerce client’s point-of-view: Every impression shown on a poor performing ad is losing them money.

We assuage these fears by educating our clients on sample size, giving them regular updates on the progress of these ad tests. But maybe that isn’t the right approach.

Maybe the better approach is to change our testing metrics.

Calculating Profit per Impression

Striking a balance between Click-through Rate (CTR) and Conversion Rate (CR) optimization can be difficult. But if we knew how much Revenue was being generated every time an ad displayed and balanced that with Cost, that would be far more valuable.

That’s where Profit per Impression (PPI) comes in.

More accurately, PPI derives Profit per Thousand Impressions.

The formula for calculating PPI is:

Profit per Impression Equation

Here at Portent, we use the ClickEquations bid management platform. Their “Text Ad Zoom” feature calculates PPI automatically:

Text Ad Zoom

But you can also calculate it manually using Google Analytics and Excel.

Note

Go to Google Analytics. Drill down to Traffic Sources > AdWords:

Analytics Traffic Sources

Then, drill down to Campaign > Ad Group. Select Ad Content from the main dropdown:

Adwords Campaigns

Export this report as CSV for Excel:

Excel ExportOnce the data is in Excel, you can strip out all rows and columns except for the Impressions, Cost and Revenue, leaving you with something like this:

In the next column over, create a space for PPI and enter the following formula:

Once you’ve applied that formula, and changed your number format to “Accounting”, you can simply copy and paste it down to the cell(s) below to get PPI for the rest of your ads:

Using Profit per Impression

Now that we’ve arrived at our PPI metric, we can make better decisions with our split-tests.

For example, take these two ads:

Both the Top Ad (A) and the Bottom Ad (B) have run on roughly the same amount of Impressions. A is the clear winner in terms of CTR. But that doesn’t factor in Conversions or CR.

Now look at Conversions and CR for each ad:

B is the clear winner in terms of Conversion and CR. But that doesn’t factor in Cost or Revenue at all. The 6 Conversions on A might’ve brought in far more Revenue than the 8 Conversions on B. Furthermore, the 106 Clicks on A might’ve come at a lower Cost per Click than the 75 Clicks on B.

So let’s bring in PPI:

Low and behold, B takes the cake. It brings in approximately $43 more per thousand impressions shown.

Profit per Impression: The Perfect Metric

Instead of relying on purely CTR, CR, or even Revenue alone to make ad decisions, PPI leverages the only constant in your basic ad split-test – equal impression amounts – and it factors in what your client cares about most: ROI.

Take the guess work out of your ad testing, and use PPI.

—————-

  • Chad Summerhill

    Hey Michael, enjoyed the article!

    Here are a couple of other related articles that your readers may find useful.

    @bgtheory:twitter
    wrote a great article on this method: http://certifiedknowledge.org/blog/profit-by-impression-the-real-metric-in-ppc-testing/

    @digitalalex:twitter

    Just mentioned PPI today as well: http://searchenginewatch.com/3642306

    http://www.wordstream.com/blog/ws/2010/09/15/ppc-ad-testing-to-maximize-conversions ( FREE spreadsheet mentioned in this post: http://www.chadsummerhill.com/ppc-text-ad-test-statistical-validity/)

    -Chad

  • Pingback: Profit per Impression: A Better Metric for Ad Testing | Iuvo

  • http://wellontop.com/ Sean Weigold Ferguson

    How would I determine if two PPI’s are statistically significantly different from one another?

    • http://www.facebook.com/MichaelWiegand Michael Wiegand

      Hi, Sean.

      I’d set a threshold of % change and go from there. 
      When you’re dealing with less money, look for a higher % of change. When you’re dealing with more money, look for a lower % of change.

  • http://twitter.com/PPCNI Jordan McClements

    Interesting article.  Though I am sceptical that 75 clicks would ever be enough clicks to tell you anything…

    • http://www.facebook.com/MichaelWiegand Michael Wiegand

      Thanks for the comment, Jordan.

      I just used an arbitrary example there. In the real world, you’d obviously balance PPI with clicks generated.

      The point of PPI is that you’ll never control click traffic on given ad test. But you *can* control what percentage of impressions your ads get.

      Hope you find the metric useful!

  • Dylansamuelwright

    Hey great article. I rarely learn something very useful from Ppc articles

  • http://www.climaxmedia.com/ Climax Media

    This is a fantastic overview of metrics to consider when running CPC campaigns and split testing ad copy. Thanks for the tips. Just used it to set up custom reporting in analytics.

Close

Sometimes we all need a little help from a hero, and that's exactly what we offer at PPC Hero. From a simple account audit to full time management, you can hire a hero to work on your accounts!