Less Is More, And More Is Less?: Calculating Real CPM When Adding Placements

When Less Is More: Measuring the Value of New Ad Placement

The conversation is inevitable. While the phrasing always varies, it starts along these lines:

“We have some extra space in the footer, do you think we could squeeze another 728×90 in there?” or maybe, “Most of our article pages extend beyond 1400 pixels, that means we can fit another 160×600 on the rail, right?”

Market inefficiencies are meant to be exploited, and the CPM ad model is easily taken advantage of. As long as more ad inventory equates to additional revenue, incentives exist to paste another placement on a page. 

The surface benefits of this ‘strategy’ are clear. In most cases adding that buried 728×90 to the footer will lead to additional impressions, increased revenue, and a higher page RPM. Even in scenarios where the unit is excluded from your direct sales, and performs poorly in programmatic auctions, you’re still monetizing 100% of the impressions and returning a measurable daily revenue boost.

So what’s the problem? Additional revenue is hard to argue with, and increasing overall earnings by pasting another display banner on your site may seem like a low hanging fruit solution, or a stimulus package to the C-Suite. 

As publishers, we understand it’s not always that simple. Having the pleasure of working at a boostrapped startup over the past 8 years, we’ve had many discussions about the number of ads we should run on a page, and the effect that has on the user experience vs. revenue gains.

For publishers dealing outside of the direct space and more in the programmatic/long tail (like myself), a balancing act between maximizing impressions and page RPM is a critical strategy for revenue growth.

During a recent re-design of one of our sites, we elected to add a 5th placement to the page vs. sticking with the 4 ads we were currently running. We weren’t kidding ourselves, we knew 5 ad placements was an aggressive play (it’s the max Google allows for participation in AdX) but we also had agreements in place with our managed demand and programmatic partners, and felt confident we could monetize the extra impressions at a reasonably high rate compared.

My team and I closely monitor our websites’ ad performance as part of our yield management process. Every day we pull data and look at performance on a per site, per ad unit, per partner level. We check for things like discrepancy, CPM increases or decreases, and any overall trends we can use to make serving adjustments in an attempt to increase yield. 

One of the metrics we calculate daily is what we call ‘rCPM’ (real CPM). We define rCPM as revenue/(total monetizable impressions/1000). We use this to account for total monetizable impressions available for an ad unit (according to the ad server or analytics program). We feel this is a more accurate measurement than eCPM because it accounts for lost impressions from bad ad calls, serving discrepancy or latency where eCPM just looks at paid impressions. It’s also a great way to look at your overall ad performance. 

Upon launching the 5th ad placement, we quickly noticed a decline in overall rCPM site wide. That wasn’t surprising because we anticipated the new impressions would be of lower value than those from the other units, pulling down overall performance. What was troubling however, was we also saw that percentage revenue growth and percentage impression growth were no longer in line as they had been prior to launching the unit. This correlation, combined with the decrease in sitewide rCPM indicated to us that the new unit impressions were actually not as valuable as we had hoped and were potentially having an impact on our other units’ performance. Our concerns were both long and short term. 

As a result, we decided to measure this two simple ways, outlined below. If you’re facing similar questions, give them a shot. While they won’t decide your next step,  and certainly wouldn’t be confused for an advanced economic model, they will provide some basic intel you can use make a more informed decision.

Option 1 – Index and measure your rCPM, impressions and revenue

  • Step 1 – Determine the date range you want to measure. You should include the launch of the new ad unit, as well as at least a few weeks of previous data to use as a performance baseline before the launch.
  • Step 2 – Calculate and index your website’s rCPM, total impressions, and total revenue from the display advertising you wish to measure. This would also work for mobile. If it’s been a while since you used index numbers, here’s a quick walkthrough.
  • Step 3 – Plot that data in Excel with the date range on the x-axis and the indexed numbers on the y-axis.
  • Step 4 – Analyze your findings and pay close attention to the impact on revenue. You want to make sure that the inevitable rise of site impressions with the launch of the new unit are providing an adequate boost in revenue. If not, you might want to consider the long term effect this may have on the performance of your other ad units and user experience.

Option 2 – Compare your revenue per thousand users before and after a new unit launch.

  • Step 1 – Determine the date range you want to use as your baseline value. Grab your website’s total unique users and total ad revenue for that period. 
  • Step 2 – Determine your per user session CPM rate. Revenue/(Sessions/1000)
  • Step 3 – With your baseline data in hand, take the date range post launch of the new ad unit and determine your new unique CPM rate. 
  • Step 4 – Assuming there’s an increase, is that increase measurable enough to justify running the additional placement? How much has the per user session CPM increased? These are important questions to consider and discuss internally before deciding on next steps.

As a result of both tests indicating a negative impact on overall site performance, we debated our next steps. Instead of eliminating the placement entirely, we elected to remove it from pages where it was most displaced from the user flow, and therefore less likely to be engaged and perform well programmatically. Even though we’d receive less overall impressions, the remaining inventory would be more valuable. Two weeks later, we re-measured performance and things had improved. It wasn’t our best performing unit but, it was now a positive contributor.

While there will always be myriad factors affecting overall ad performance, the more you’re able to understand about placements and how each individual unit compliments one another, the more prepared you’ll be to discuss next steps or build a case for or against an additional ad.