We all want more for less. Whether you’re an online retailer promoting your products or a publisher trying to get your articles in front of the right eyes, we all want to pay as little as we can while driving as much traffic as possible to our sites. This is especially true for those in the content business, where content is plentiful but conversion rates tend to be low.
One of the primary challenges content marketers face is making sure that they’re getting the best cost per click that they can for their Facebook campaigns. While we have yet to solve all the challenges of being a content marketer, we have arrived at a place where we can tell content marketers which CPC they should be using.
You wouldn’t want to spend your entire campaign budget at a CPC of $0.50 when it’s possible to pay $0.05, right? To avoid situations like that, we use a process called sampling, which helps us know if we’re getting the lowest CPC that we can for our ads and not compromising on audience quality.
To be clear: When running Facebook ads, Facebook actually does the sampling for you to determine the best ad in an ad set. The challenge arises when comparing different messages with different audiences, since Facebook ad sets can only target one specific audience apiece.
This challenge is particularly acute for content marketers, since they have numerous (hundreds or even thousands) of articles to promote to different audience and, as a result, hundreds or thousands of different ad sets. We’ll get into that a bit later, but for now, a high-level understanding of sampling will help frame the discussion.
Crash course in sampling
Sampling is much like your standard run-of-the-mill A/B testing. To illustrate this, let’s compare sampling to–stay with me here–buying a car. You don’t need to take a cross-country trip to determine whether you like it, but you should spend enough time behind the wheel to feel confident making the big purchase.
In this sense, sampling an ad is much like test-driving a car; it allows you to gather preliminary information on how an ad is likely to perform.
Also, you shouldn’t limit your car hunt to the first one you take for a spin. So when sampling, rather than running an entire campaign trusting a single set of creative and target audience, you test multiple versions of an ad and target a variety of audiences. From there, you can use the insight garnered from sampling to inform how to move forward with your campaign.
Consider this simple scenario: Mike is a content marketer with a total budget of $100 for his campaign. He wants to get the best CPC that he can, so he samples two versions of his ad and gives them each a budget of $0.50.
- Version A is shown to 50 users and five people click on it, giving him a CPC of $0.10.
- Version B is also shown to 50 users and 10 of them click on it, giving him a CPC of $0.05—one-half the cost of Version A.
Seeing that Version B’s CPC is much lower, he goes all-in and allocates the remaining $99 from his budget to Version B.
Unfortunately, Mike’s approach to sampling is flawed: 50 impressions are too few to predict how the ad will perform long-term, and the difference of five clicks could have happened simply by chance. More data is needed to predict how an ad will perform.
At the same time, Mike doesn’t need to test the ad for 50,000 impressions–that strategy would be too costly in terms of both time and money, making it unsustainable.
So where do we draw the line?
The next step in sampling is to figure out how much sample data is needed to establish that the data will be representative of the lifetime of the ad. Given the volatile nature of an ad’s CPC at the beginning of its run, you want to avoid making a judgment call too soon in the event that it will change later on–like in Mike’s case. You also don’t want to wait too long and waste your budget on a poorly performing ad that could be better spent on one that’s more effective.
Therefore, when sampling ads for a campaign, as with most things in life, it’s important to know when to stop. Unsurprisingly, this is known as the sampling stopping point. This is the earliest point at which you’ve gathered enough data to safely predict the ad’s overall CPC. The CPC observed through the sampling stopping point will closely resemble the CPC for the lifetime of the ad, so beyond this point there won’t be much fluctuation in CPC.
Take the following hypothetical example:
Here, the sampling starting point was probably somewhere around 60 or 70 clicks into the campaign, which is where the variation in CPC began to level off. As previously mentioned, Facebook’s ad platform automatically does this for you. After a short sampling period, Facebook’s algorithm chooses the best ad in an ad set and suppresses the others that did not perform as well.
When to stop
Sampling stopping points can vary greatly depending on a number of factors like industry, campaign objective, audience size, Median CPC price point and Facebook’s pacing algorithm that is affected by the budget and duration of the ad. For CPC campaigns, sampling size can range from $1 to tens of dollars. We generally reach a sampling stopping point at about 100 impressions and an ad spend of $1.20.
While sampling is very useful, the time required to manually repeat the process can become a major drawback. The person managing the ads has to perform the calculations over and over, keeping a constant eye on the ad dashboard.
The fact that you can only target one audience per ad set makes this an even more arduous task to undertake. While you’re able to compare how different versions of an ad will perform within an ad set, the results will only be applicable to the singular audience being targeted. To compare across different audiences, marketers must create separate ad sets for each audience, which can quickly become cumbersome and disorganized.
These factors can make it difficult to scale sampling at a high volume when the primary variable is the audience. Imagine being responsible for audience development at a major publication: You have plenty of content available at your fingertips and virtually endless possibilities for target audiences. Given the multitude of options this makes possible, it would be tough to know that you’re reaching the right audience while still paying as little as possible for your CPC.
Fortunately, there are automated tools in the market today that address this issue.
With sampling, you’re able to test-drive variations of an ad to uncover the most cost-effective version to use for your campaign. The point at which you’ve gathered enough information that you can comfortably predict how each ad will perform is the sampling stopping point, which is typically 100 impressions or a $1.20 total spend (the higher of the two).
Yaniv Makover is the co-founder and CEO of Keywee, a platform that enables the creation, distribution and measurement of content that drives business results for publishers, retailers and brands.
CPC image courtesy of Shutterstock.