DISCLAIMER: This is my interpretation and oversimplification of how Facebook (and many other ad platforms) work, based off my own personal experience ($25m spent across channels) and countless conversations with account reps and other marketers.
NOTE: You will hear conflicting reports even from different people within Facebook as to how all this works. The truth is; it’s really complex, most account reps don’t know, most developers don’t know, (#MachineLearning), and algorithms change all the time.
The best approach is always to:
- think logically about whether the advice someone is giving you even makes sense based on Facebook’s incentives,
- test it and see if it works to improve performance, then
- update your assumptions with what you learned (and we'll keep doing that in this post as we learn more!)
Forget what you, as a marketer or an advertiser, want for a second. What does Facebook want?
The answer: to make more money without pissing off users.
So here's a simplified calculation of how that might work:
$$$ per User = CPM * # of Ads.
Let's break it down:
Download a PDF version of this post!
# of Ads
If Facebook’s users get pissed off and leave, they can’t show them ads anymore. For that reason, Facebook has to strike a delicate balance between providing engaging content from friends and sponsors in order to avoid overwhelming users with ads.
That's part of the reason why relevance score exists.
Relevance score = engagement / complaints
If users hate your ad, or don’t engage with it, Facebook's advertising algorithm will also hate you and will likely either charge more money or just reject showing your ad to their users.
Note: this serves the same purpose as Google’s ‘Quality Score’ rating.
Keeping an eye on your relevance score as you run ads is important to make sure you're not wasting money on an ad that your audience clearly doesn't appreciate.
CPM, or cost per mille, is a cost calculation of how much money it would cost to land 1,000 impressions on your Facebook ad.
If Facebook does a better job at finding the right users, advertisers are willing to pay a higher CPM to show ads (impressions) to them. That's why their team works so hard on expanding targeting options and increasing efficiency of their advertising algorithms.
Here's how the math works here:
$50 Conv. Value * 2% CVR * 1% CTR = $10 CPM
- Conv. Value = Revenue per Conversion
- CVR = Conversion rate
- CTR = Clickthrough rate
- CPM = Cost per Mille (1,000 impressions)
$50 Conv. Value * 2% CVR * 2% CTR = $20 CPM
If Facebook matches you to users that click twice as often, they can earn a 2x higher CPM.
$50 Conv. Value * 4% CVR * 1% CTR = $20 CPM
Same goes for conversion rate. Double the conversions means double the CPM they can charge.
Note: this is only if advertisers are efficiently bidding. There are some caveats and inconsistencies that allow platforms to game the system, but over the long run this holds true.
To decide what users to match you to, Facebook needs to know what your objective is. That can be anything from clicks to conversions to impressions... there are plenty of options Facebook provides you depending on the type of ad you want to run.
If it’s clicks, they’ll match you to people who historically show a high clickthrough rate on ads. If it's conversions, they’ll look for people likely to convert, based on what they’ve seen on your pixel so far.
The more data they have, the better they can match your ad with the exact type of user that would convert. Rule of thumb here is 20 conversions in 30 days to get the best match.
But how exactly do they match you to the right users?
Say we only had 10 users in our Facebook audience, each with a different likelihood to convert and some more expensive to reach than others.
Other advertisers are trying to reach these users too, maybe based off different targeting than what you’re using.
So the first thing they’ll do is order these users descending by likelihood to convert (or click, if that’s your objective).
Next, they’ll figure out which are the cheapest to reach. Let’s choose the cheapest 40% of the audience we have.
When Facebook shows your ads to these people, they know they’re giving you the best possible chances of success while still maximizing your budget usage.
The real algorithm is much more complex, but it’s the same basic idea:
- find the best users in your audience
- do it for the cheapest price
- maximize the effectiveness of your budget
- and keep you coming back for more
You’ll notice that Facebook doesn’t just blindly minimize my CPM: I’m paying $2 for Dave, because he has a better propensity to convert than any of the cheaper options below him.
I completely skip Sarah and Sally: Facebook can now sell those impressions to other advertisers that they know are more likely to convert and will be willing to pay a much higher CPM.
That’s how Facebook makes more money while also saving their advertisers money.
Of course they do all this at lightning speed with complex machine learning algorithms based off millions of variables, but you get the idea.
If you’re a smart advertiser, you’ll notice that improving your conversion rate, revenue per conversion, clickthrough rate or targeting has a direct impact on your bottom line.
That’s why it’s important to test new ad creative (CTR / Relevance score), landing pages / onboarding (Conv. Rate), and CRM campaigns (revenue per user).
You can either take the cost savings and bid down on CPM to be more profitable, or you can bid up on CPM to get more volume (attract more of the Daves and Bills of the world).
You might not know this, but every Facebook audience is a lookalike audience. Even if the example above was “Millenials who like Dogs and buy Pet Food in Manhattan”, Facebook has just taken the top 20% of that audience and shown your ad only to them.
It’s why you almost never reach the full audience, and why it hurts performance to increase spend -- everyone else is either too expensive or not as interested in you.
Lookalike audiences work exactly the same way, but with the entire population of a country. Ever noticed how a 1% lookalike of any conversion event or email list for any client always equals the same number of people?
It’s because they’re ordering the entire country based on their likelihood to convert on your pixel, then selecting just the top 1% to go into your audience.
This approach is wildly successful and performs amazingly well because Facebook’s algorithm is better than a human at determining “likelihood to convert”. Further, it finds these cheaper people you wouldn’t normally find.
For example, maybe you know that people who use your product are Entrepreneurs who like Gary Vaynerchuk, use Mailchimp, are 25-35 years old, and are living in New York.
Source: Startup Hubs
BUT do you know that there are tons of small pockets of entrepreneurs living in places like Texas and North Carolina? Do you have a list of every single startup influencer or email service provider? Can you identify the occasional 45 year old who would also like your product?
Facebook can do all of the above. They have several orders of magnitude more data than you do. More importantly, they can decide what’s more relatively important to predict whether someone will convert.
That’s why lookalike audiences work so well, why Facebook is one of the best ad platforms ever created, and why we still use them daily, despite all the hours wasted dealing with bugs and inadequate customer support.
Download a PDF version of this post!
Choice of Objective
Facebook needs data to do what it does best. So if you choose a conversion pixel too far down the funnel, it won’t find the right people.
If you choose an objective too high up the funnel, like clicks, it’ll optimize to people likely to take that action, regardless of whether they’re likely to convert further down the funnel.
Unfortunately, clicks do not equal conversions.
Sometimes I’ve seen click campaigns drive much better CTR on the same audience, but at a much worse cost per acquisition because the conversion rate was so awful. And that doesn't even consider revenue-driving factors like qualified vs. unqualified leads, average order size, lifetime value, etc...
Here's how each objective works:
$20 CPM * 0.5% CTR * 2% CVR = $50 CPA
$20 CPM * 2% CTR * 0.2% CVR = $125 CPA
Changing the campaign objective should not affect reach of your campaign. All it affects is Facebook’s ability to choose the right people from your audience.
By default everyone agrees to spend the full daily budget, regardless of performance, and to pay Facebook by impression.
There is zero risk to Facebook getting paid if they show your ads, so as long as your ad relevance score is good enough to show you in the first place (2-3+), you’ll spend your budget.
What does affect the reach? Target bids.
Rather than choosing ‘maximize conversions’ you can actually set the price you want to pay per conversion. In fact, you can go one step further and refuse to pay Facebook at all if they don’t drive a conversion.
Of course this is great for you - no risk whatsoever. Only pay if you make a sale. The ability to pay for clicks rather than just showing an ad with no guaranteed traffic, is part of why Google Adwords was successful.
However there’s a downside: reach.
Facebook can just decide not to show you. And can you blame them? They aren’t going to get paid if your ad doesn’t get clicks, or if it doesn’t convert.
Why take the risk when they have literally millions of other advertisers who go with the default of ‘spend all my budget, whatever happens’.
This used to be a very popular option, but over time I’ve noticed it’s largely stopped working for all but really large advertisers where Facebook has enough data on whether a user will click and convert for them.
Even still, it’s likely a drain on reach for those advertisers - they could be getting more volume if they were less conservative and took more risk.
As you can imagine, some hours, days or weeks are more expensive on Facebook than others. If Disney is launching a new movie, you can imagine they’re spending millions buying up all the inventory, increasing the prices for everyone else.
So one way to improve performance is to move to a lifetime budget. This takes away some control in how much you spend per day, and some more fiddling around when the campaign ends, but it lets Facebook spend less on expensive days, more on cheap days for you.
Default to conversion campaigns. Note that you can’t change this afterwards (i.e. from Traffic to Conversions), and you can optimize to clicks or impressions with a conversion campaign anyway.
Facebook learns from all your organic traffic converting on your pixel.
Optimizing to clicks will get you a lot of shitty clicks. You should have a conversion goal for every campaign (driving awareness is worthless)
Default to higher up the funnel. To get more data, try optimizing to the first step of your funnel. Don’t go too far (clicks) because you don’t want poor quality traffic.
Watch your funnel like a hawk - optimize each stage to improve your ROI.
Learn how to do these funnel calculations when planning campaigns.
Understand that the more volume you need, the more expensive your campaign will be.
Don’t assume that ads with high CTR will also have high CVR.
Don’t use CPA caps / target bids unless you’re willing to spend nothing.
Use Lifetime Budget if consistency of daily spend not an issue.
Grow your business with Facebook ads, conversion rate optimization, and full-funnel marketing.