In marketing, what is a CPA?

This sounds like one of the most introductory questions to marketing that you could ask, but I think very few people fully grasp what a CPA represents. As easy as the answer to this question might seem, understanding the answer at a fundamental level can teach us a lot about how to run efficient marketing.

So, what is a CPA?

CPA typically stands for cost per action. You might see it used as an abbreviation for something else, like cost per acquisition, but for the purpose of this article we're going to treat all these variants as interchangeable.

A CPA is a measure of how much it costs you to drive a particular action. This action could be a purchase on an ecommerce site, a registration to your newsletter, or whatever action you care about in your marketing.

It's calculated by taking your total marketing spend, and dividing it by the number of actions you've driven with that spend.

To give an example, let's say you've spent $10,000 on a channel, and driven 1,000 actions with that channel. Your CPA would be $10,000 / 1,000 = $10.

So what's the big deal?

Assuming that you're happy dividing one number by another you're probably asking yourself at this point: what's the big deal?

The big deal, so to speak, is in recognising what precisely a CPA represents. Because a CPA is calculated by dividing the total spend by the total actions, a CPA is an average. That is, a CPA figure for a channel represents the average cost per action for that channel.

The problem though, is that averages are misleading. A river that's 4 feet deep on average isn't necessarily safe to cross; certain points could be far deeper than 4 feet, even though the depth averages out at 4 feet.

In just the same way, the fact that a channel's CPA is $10 doesn't mean that each action driven by that campaign costs $10.

How much does each action cost?

Let's say you run the example campaign that we looked at earlier, where you drive 1,000 actions on a channel for $10,000 worth of spend. What will happen is that:

• The first hundred actions are the cheapest to drive. This is because very few people on this channel have seen your brand yet, and you can concentrate your initial spend on the people that your brand is most likely to resonate with.
• The last hundred actions are the most expensive to drive. By this point, many people on the channel have seen your brand (and so already made up their mind about whether to convert on your ads), and so you have to spread your budget to reach lower-intent users, who your ads have less chance of resonating with.

Here we start to get a sense that, in a campaign with a $10 average CPA, not every action actually costs $10. Some will cost much less, and some much more.

Something important to note about the above is what I mean when I speak of the first or last hundred actions. When I refer to the first hundred actions, I'm not referring to the first 100 actions that take place in time, since the start of the campaign.

Instead, I'm referring to the idea of running a campaign that drives only 100 actions, in comparison to a campaign that drives 1,000 actions. A 100 action campaign is likely to saturate a channel much less than a 1,000 action campaign, and so the cost per action on the 100 action campaign is likely to be much lower.

Similarly, when I speak of the last 100 actions, I'm drawing a comparison between a campaign that drives 900 actions and a campaign that drives 1,000 actions. Let's say that the campaign which drives 900 actions has started to saturate, and so the cost of an extra 100 conversions will likely be much higher than $10 each.

One way of simplifying this talk is to think about the potential costs and volumes of the campaign as sitting on a curve:

‍

This is a cost-volume curve. On the x-axis you have the number of actions driven the campaign, and on the y-axis you have the cost of that campaign.

The first thing you'll likely have noticed about the curve is that it isn't straight, i.e. it's a curve. The fact that it's a curve is equivalent to our statement earlier that not every action costs $10:

• On the left hand side of the curve, where volume and spend are low, actions cost below $10 each.
• On the right hand side of the curve, where volume and spend are high, actions cost more than $10 each.

The cost of an individual action is actually linked to the slope of the curve at that point. If you're planning a campaign to drive 250 actions, and you want to find out how much that 250th action costs you, you can find out by taking the curve of the slope at that point.

• Mathematical explanation
  ‍The curve of the slope is it's gradient, and reflects how a change in the x-axis variable leads to a change in the y-axis variable. In this case, the gradient reflects how a change in number of actions (i.e. generating 1 more action) leads to a change in spend. This is the answer to how much more do I have to spend to get 1 more action at this point?

Now we understand the relevance of the curve's slope, we can use this to understand what we said earlier about why the cost of the first 100 actions is lower than the cost of the last 100. The fact that the slope of the curve is low on the left hand side reflects the lower cost per action at this point on the graph, and vice versa for the right hand side.

When we talk about cost per action or CPA at a particular point on the graph though, there's a specific term we should use: marginal CPA. A marginal CPA is how much it costs us to get one more action at any point on the graph, and is equal to the slope of the curve.

Why does this matter?

It's easy to look at all this and think; cool, but why should I care?

Sadly I think the large majority of marketers overlook the importance of marginal CPA as a metric. I would go so far as to argue that it's far more important than a regular old average CPA, as far as metrics are concerned.

I have another article which outlines the reasons for this in much more detail, but I'll give a quick run-down here.

Let's say you're selling a product with a $30 profit margin. What CPA should you aim for?

If you're just using CPA as a metric, you won't have much luck answering this question. You know the answer should be less than $30 (or you won't make any gross profit), but how far below? A $20 CPA? A $10 CPA? Who knows!

If you're focused on marginal CPA, there's a dead easy answer to this. Your marginal CPA should be $30 or less. If your marginal CPA is below $30, then you can be sure you're making some amount of gross profit on every sale that your marketing generates. It doesn't matter what your overall CPA is, as long as your marginal CPA is $30 or below, you'll be in the money.

So, how do I find my marginal CPA?

Finding out what your marginal CPAs look like is the last hurdle, and it requires you to build up a cost-volume curve like the one we looked at earlier. There are a couple of ways to do this:

• You could use tools built into ad platforms, like Google Ads' bid simulator, to help you determine what your cost and volume data points are.
• You could test running channels at different spend levels, and make a note on how volume reacts to changes in spend.
• You can segment your performance data at a granular level (i.e. ad level) and use the cumulative approach outlined here under Beyond bid simulator.

Any of these approaches will work, and let you get a handle on what the marginal CPA curves look like for your marketing. Definitely take the time to do this; most advertisers are often surprised by just how much higher their marginal CPAs are above their average CPAs.

• Before you go...
  If you've made it this far, and you're interested in marginal CPAs, I'd genuinely love to hear from you. Marginality is an under-explored topic in marketing, and I really want to learn how marketers are approaching and thinking about it currently. In return, I'm happy to share my thoughts and be helpful in any way I can.
  
  I'd love for you to contact me here.