Relying On Averages Is Bad Business
Too many people rely on stats that purport to measure “averages” to guide their search marketing campaigns, and they are all potentially harmful. I am the opposite—you could say that I am passionate about accuracy when it comes to analytics and when it comes to all forms of measuring your marketing campaigns. In this article […]
Too many people rely on stats that purport to measure “averages” to guide their search marketing campaigns, and they are all potentially harmful. I am the opposite—you could say that I am passionate about accuracy when it comes to analytics and when it comes to all forms of measuring your marketing campaigns. In this article I will outline scenarios where being a stickler for detail can have a large impact on the success of your campaign. The focus of the scenarios outlined will be on pay per click (PPC) campaigns, but the general concept of striving for the highest possible accuracy applies to any type of marketing effort.
Below, a look at common measures that can have a potentially harmful effect on a search marketing campaign if not used appropriately.
Cost per action (CPA) vs. return on ad spend (ROAS)
One of the popular methods for tracking results is to calculate the cost per action (CPA). Basically, with this method you calculate how much it costs you for each action generated on a per keyword, per ad group, or per ad campaign basis. This method works great when the value of each sale is exactly the same.
Many companies, or their agencies, use the CPA model even when the revenue event is not identical in every case. In principle, this concept can actually work for you. Let’s say you know that your average sale online is $50, and you are willing to spend half of that on a sale. You could set your CPA goals to $25 and manage to achieve those goals and probably be OK.
However, you are most likely leaving some money on the table in this scenario. We have seen clients that have great variability in the value of each sale who were running their campaign in a CPA model. They were, in fact, doing OK with this approach to their search engine marketing campaigns.
The problem is that the sales do not have equal value. You could have one product that sells for $200, and a bunch of products that sell at $25. If one of your key phrases is driving the $200 product sales in a big way, why handicap it with a CPA of $25? Clients of ours that have adjusted from CPA to ROAS have seen significant gains.
Revenue vs. margin (or ROAS vs. MOAS)
I’d venture to say that nearly every search marketer is in the habit of tracking revenue in PPC campaigns and using that as the basis of adjusting bid prices. In general this is OK, provided that you are setting your ROAS goals that provide enough return so that it provides you with the desired volume and profit from your campaign.
But, once again, the law of averages leaves money on the table. Even if all your products sell for $50, some of them may have 10% margin, and others may have 70% margin. For exactly the same reasons you want to focus on ROAS instead of CPA, you may want to focus on margin instead of revenue.
It’s actually usually quite easy to do. In any ROAS based system, you end up implementing some web application that plugs the correct revenue number into an analytics script at the completion of every transaction. Well, instead of plugging in the revenue number, try plugging in the actual margin on the sale. This is sweet fine tuning in action!
Cost Per Lead vs. Actual Revenue (again)
There are plenty of businesses that don’t complete their transactions on site. For example, in the financial services industry a phone conversation is almost always necessary to close a sale. Yet these types of businesses can develop lots of business from the web. They just need to use a cost per lead model, and this works reasonably well for people lacking other options for measurement.
However, like all estimating techniques, there are hidden lies in the cost per lead model. Recently, I was speaking to Adam Goldberg of ClearSaleing, and he provided me with 2 specific examples of this. The first one is an auto loan client of theirs:
As you can see in the example, the cost per lead of the two campaigns is almost identical, but the refinance campaign produces more than 20 times as much revenue as the new loan campaign. There is a clear reason to treat these differently. Of course, once you realize this, you can use a different cost per lead model for both campaigns. But let’s look at another example:
In this one a cost per lead model would be telling you to shut your Google campaign off and scale up your Yahoo campaign. This would actually be disastrous, and quite the opposite approach is what should be taken—bids on Yahoo should be tweaked to get it to be profitable, and the Google campaign should be scaled.
Clearsaleing is a tool that allows you to reconcile actual transaction data with your marketing spend online. As you can see, this can have quite a dramatic impact.
We do need to make best guesses in a lot of what we do. It’s a fact of life, and we should not shy away from it as long as the approach is mathematically sound. But, when you can, eliminate those sources of error and you will usually be well rewarded. Better still, since many of your competitors may not put in that extra effort, this could create a competitive advantage for you.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.