Welcome to the second installment in my Building the Perfect Display Ad Performance Dashboard series (Note to self: pick a shorter title for the next series). In the first installment, we looked at an overarching framework for thinking about ad monetization performance, comprised of a set of key measures and dimensions. In this post, we’ll drill into the first of these – the measures that you need to be looking at to understand your business.
How much, for how much?
As we discussed in the previous post, analysis of an online ad business needs to focus on the following:
- How much inventory was available to sell (the Supply)
- How much inventory was actually sold (the Volume Sold)
- How much the inventory was actually sold for (the Rate)
Of these, it’s the last two – the volume sold and the rate at which that volume was sold – where the buck (literally) really stops, since these two combine to deliver that magic substance, Revenue. So in this post we’ll focus on volume sold, rate and revenue as the core building-blocks of your dashboard’s metrics.
Volume, rate and revenue are inextricably linked via a fairly basic mathematical relationship:
Revenue = Rate x Volume
Another way of thinking about this is that these three measures form the vertices of a triangle:
Some business and economics textbooks call Rate and Volume “Price” and “Quantity” (or P and Q), but the terms we’re using here are more common in advertising.
Different parts of an ad business can be driven by different corners of the triangle, depending on the dynamics of how each part is transacted. Here are some examples:
- Ads sold on a time-based/”sponsorship” basis are best thought of as driving revenue performance, because deals are done on a revenue basis regardless of volume/rate (though the advertiser will have a volume & rate expectation, which they’ll want to be met).
- For premium ads sold on a CPM basis, deals revolve around Rate; the name of the game is to add value to inventory so that, impression-for-impression, it achieves more revenue.
- For remnant ads and networks, volume is king (assuming you can maintain a reasonable rate) – you’re looking to maximize the amount of inventory sold, and minimize the amount that has to be given away or sent to “house” advertising.
Because of these different dynamics, measurement of ad monetization can easily fragment into various sub-types of measure; for example, as well as cost-per-thousand (CPM) rate, some ads are purchased on a CPC or CPA basis. So a more complete version of the diagram above looks like this:
However, it’s essential to remember the key relationship and dynamic between rate, volume and revenue, which is manifested in the CPM, Impressions and Delivery Revenue measures in the diagram above. So let’s look at these measures.
In the online ad business, Volume is measured in Ad Impressions. I have talked about ad impressions before on this blog, in this installment of Online Advertising 101 (you may want to take a moment to read the section entitled “What’s the product?” in that post). From a measurement point of view, whenever your ad server serves an ad (or more accurately, fields a request for an ad), its measurement system should log an ad impression. How much data is logged with this impression will vary depending on the ad server you’re using, but will likely include most of the following:
- Date & time of the impression
- Campaign and/or creative
- Location/placement (i.e. where the ad was served)
- Attributes of the individual who requested the ad (e.g. targeting attributes)
We’ll come back to those attributes (and how you can use them to segment your impressions for better analysis) in another post.
Capturing a true view of the ad impressions on your site can be a little more challenging if you are using multiple ad servers or networks to sell your inventory, particularly if you are using a combination of your own first-party ad server (for example, DFP) and redirecting some impressions to a third-party such as an ad network. When you have delivery systems chained together in this way, you may need to combine the impression counts (and other data) from those systems to get a true picture of impression volume, and you will need to be careful to avoid double-counting.
For reasons that will become clearer when we get on to talking about rate, it’s essential that you capture impression counts for your ad sales where you possibly can, even for parts of your site or network where the supply is not sold on an impression basis.
Other volume measures such as Clicks and Conversions become very useful when you’re looking to assess how valuable your inventory is from an advertiser perspective, since both are a proxy for true Advertiser ROI. They’re also useful for deriving effective rate, as we’ll see below.
At the highest level, rate is a simple function of volume and revenue – simply divide your total revenue by your total volume (and usually multiply by 1,000 to get a more usable number) and you have your overall rate – in fact, you have the most commonly used kind of rate that people talk about, known as “Effective Cost-per-Mille (Thousand)”, or eCPM (don’t as me why the e has to be small – ask e.e. cummings). Just to be clear, eCPM is calculated as:
eCPM = (Revenue) * 1000 / (Volume)
Sometimes eCPM is known as eRPM (Where the R stands for “Revenue”).
The reason we’re talking about eCPM before revenue in this post is because many advertising deals are struck on a CPM basis – i.e. the advertiser agrees to buy a certain amount of impressions at a certain pre-agreed rate. However, even for inventory is not being sold on a CPM basis, it’s essential to be able to convert the rate to eCPM. Here’s why.
The beauty about eCPM is it is the lowest common denominator – regardless of how a particular portion of your impression supply was sold (e.g. on a cost-per-click basis, or on a
“share of voice” or time-based basis), if you can convert the rate back into effective CPM you can compare the performance of different subsets of your inventory on a like-for like basis. Consider the following example of delivery info for the parts of a fictional autos site:
|Site area||Sold as…||Deal|
|Home page||Share-of-voice||$10,000 up-front|
|Car reviews||Reserved CPM||$2.50 CPM|
With just the information above, it’s impossible to understand whether the Home Page, Reviews or Community site areas are doing better, because they’re all sold on a different basis. But if you add impression counts (and, in the case of the Community area, click counts), it’s possible to derive an overall rate for the site, as well as to see which parts are doing best:
|Site area||Sold as…||Deal||Impressions||Clicks||CPC||Revenue||eCPM|
|Home page||Sponsorship||$10,000 up-front||5,347,592||n/a||n/a||$10,000||$1.87|
|Car reviews||Reserved CPM||$2.50 CPM||3,472,183||n/a||n/a||$8,680.45||$2.50|
See? Who knew that the Community area was throwing off so much money per impression compared to the other areas?
eCPM isn’t the only rate currency you can use, though its connection to both volume and revenue puts it at a distinct advantage, and it means most to publishers because it speaks to the one thing that a publisher can exert (some) control over – the volume of impressions that are available to sell.
If you sell your inventory on a fairly straightforward CPM or CPC basis, then your site’s revenue will pop neatly out of the equation:
(Revenue) = (eCPM) * (Volume) / 1000
However, if you’re running a larger site and engaging in sponsorship-type deals with advertisers, your revenue picture may look a little more complex. This is because “sponsorships” (a term which covers a multitude of sins) can contain multiple revenue components, some of which can be linked to ad delivery (and which therefore lend themselves to rate calculations), and some of which cannot.
For example, the sponsorship deal on our fictitious autos site referenced above could in fact contain the following components on the invoice sent to the advertiser or agency:
|100% Share-of-voice rotation, 300×250, Home Page (1 day)||$6,000||3,000,000|
|100% Share-of-voice rotation 120×600, Home Page (1 day)||$4,000||3,000,000|
|Sponsor branding – Home Page background (1 day)||$8,500||n/a|
|Sponsored article linked from Home Page (1 day)||$3,500||n/a|
|Sponsor watermark on Home Page featured video (1 day)||$1,500||n/a|
In the above table, only the first two items are expected to be delivered through the ad server; the other three are likely to be “hard-coded” into the site’s CMS and actually deliver with the page impressions (or video stream, in the case of the last one).
There are a couple of different options for dealing with this second kind of revenue (which we’ll call “non-delivery” revenue) which can’t be directly linked to ad impressions. One is to attribute the revenue to the ad delivery anyway, kind of on the assumption that the ads “drag along” the other revenue. So in the above example, with 5,347,592 impressions delivered across the two units, the “overloaded” eCPM for the ad units would be $4.39.
The challenge with this approach is that the extra revenue is not associated with delivery of any particular ad. So in the above example, if you wanted to calculate the eCPM for just the 120×600 unit on the home page (perhaps across an entire month), would you include the non-delivery revenue? If yes, then how much of it? 50%? 40%? The lack of ability to truly associate the revenue with ad delivery makes these kinds of calls incredibly hard, and open to dispute (which is the last thing you want if you are presenting your numbers to the CEO).
The other approach is to treat the “non-delivery” revenue as a separate bucket of revenue that can’t be used in rate calculations. This keeps the data picture simpler and more consistent on the “delivery” side of the house, but you do end up with an awkward block of revenue that people are constantly poking and saying things like “I sure wish
we could break that non-delivered revenue out a bit more”.
A complicated relationship
Once you have your arms around these three core measures, you can start to see how they interact, and there lies the magic and intricacy of the dynamics of selling display advertising. The implacable logic of the simple mathematical relationship between the three measures means that if one changes, then at least on of the others must also change. Only by looking at all three can you truly understand what is going on. We’ll dig into these relationships more in subsequent posts, but here’s a simple example of the rate achieved for ads sold on a fictional news site home page:
Someone looking at this chart may well ask “OMG! What happened to our rate in June 2009?” Well, a quick search on Wikipedia will reveal that a certain “King of Pop” died in that month, sending the traffic (and hence the ad impression volume) of most news sites sky-rocketing. In our fictional home-page example, almost all revenue is driven by “share of voice” (time-based) deals, so all that extra volume does is depress the effective rate, because the site earns the same amount per day regardless of traffic levels. So here’s volume and revenue from the same data set, to round out the picture:
We can now see that in fact, June wasn’t a bad month for Revenue; it was the huge spike in traffic that did the rate in.
The above example takes something very important for granted – namely, that we have enough segmentation (or “dimensional”) data associated with our measures to be able to break down site performance into more useful chunks (in this case, just the performance of the home page). In the next blog post, we’ll look at some of the most important of these dimensions. Stay tuned!
2 thoughts on “Building the Perfect Display Ad Performance Dashboard, Part II – metrics”
Another great post – thanks for putting in the time, Ian!
Lies, Damned Lies…: Building the Perfect Display Ad Performance Dashboard, Part II metrics
Comments are closed.