It's time to stop using CPMs (and CTRs) to gauge anything on Meta
If you hear this excuse – "CPMs are high [and therefore ads were more expensive today" – then stop listening.
Back in the mid-2010s, I remember hearing someone say in a meeting “oh the CPMs were high yesterday.” It was a marketer telling a non-marketer why CPAs had gone up.
Bullshit, was my gut response, that sounds like an excuse.
If you’ve ever spent considerable time in Ads Manager, you’ll know how frustratingly unreliable everything seems. High CPMs in one campaign but great CPAs. High CPMs another day with high CPAs. Then low CPMs elsewhere and CPAs all over the place.
CPMs bare very little value in how you manage your ad account. Here’s why.
There is no correlation between CPMs and CPA on Facebook
What are we looking at? Those numbers are the ‘correlation coefficient.’ A statistical measure of the strength of a linear relationship between two variables.
A score of 1 means a perfect positive correlation. And -1 means the opposite: perfect negative correlation.
Scores between +/- 0.7-1 are considered highly correlated.
Scores between +/- 0.3-0.5 are considered lightly correlated1.
Scores between 0-0.3 have little to no correlation.
In the above examples, where the is over £5m of ad spend being analysed, these coefficients fluctuate wildly.
So what can we take away from that above data – of which over £5m in ad spend has been analysed from 6 brands.
Only one account shows ‘light’ correlation – and that is only looking at an all campaign view, when aggregating daily, it’s less clear
Most accounts differ depending on if you look at daily vs campaigns.
Sometimes it’s positive, and sometimes it’s negative.
In short, there’s not a correlation you can draw from these.
The ‘highest CPM period of the year’
Excluding outliers, there is a general upward trend of overall CPMs as you get into Q3-Q4. There are two periods generally recognised as the “highest CPMs” of the year. One is Black Friday and the other is the Christmas period.
So if these are the highest CPM periods of the year, and CPMs are useful to look at, why are you advertising then?
High CPMs often signal high customer intent
Black Friday is a huge sales opportunity, and for every client that does discounting. Not just a great sales opp, but often one of the best periods of the year.
Yes CPMs go up, but so too does conversion rate and your ability to spend. People are on the lookout for a discount. People hold off buying until that weekend. Every advertiser pushes spend. CPMs rise. But your CPAs are usually good.
Christmas is the same. Loads of brands do the majority of their sales volume in Q4. And this is when your CPMs are at their highest. Doesn’t matter, customer intent is there.
Conversely, for anyone who has ever tried advertising alcohol in Jan (🙋), you’ll know you’re basically throwing your money into a big pit. But why? Most big advertisers have pulled their spend, CPMs are often lower. But the customer intent isn’t there.
Click-through rate isn’t a good proxy for practically anything
A few years ago, I was sat in one of the Facebook accelerator programmes for startups. The above chart was shown, which highlights the <1% correlation between click-through rate and ad recall, brand awareness, or purchase intent. This was Facebook’s own data. It reinforced something I was seeing in our data at the time.
And anyone who has ever done a traffic or link click campaign will be able to attest to this. Cheap clicks and high clickthrough rates don’t matter. But it’s not even just these low-value campaign types, it’s also the value of CTR amongst your conversion/purchase-focused campaigns too.
With those above brands mentioned, what’s the correlation coefficient between CVR and CPA?
-0.06, 0.01, -0.22, -0.02, 0.21 and -0.38
Only one of the accounts had a weak correlation between click-through rate and CPA. The majority had even less correlation than CPM.
Your conversion rate is not always equal
If you go and pull the data like I’ve been doing above, you’ll notice something else. Your conversion rate jumps around a lot. While it may average at 4%, there will be ads, ad sets, campaigns, days, weeks, or months when it can be 0.4% and others at 14%.
The CPM conversation assumes that if Meta places an ad that costs £6.50 CPM, and another at £2.5 CPM, then the £2.50 CPM one will have a lower CPA because your conversion rate is 4% on average.
Meta uses its data to find for you the person who, at that time, in that placement, is most likely, to do the desired action.
Sometimes, that CPM might be low, sometimes it might be high. But it doesn’t matter because your CVR isn’t always equal.
All-time FB Ads expert and influencer, Barry Hott – who like me rallies against CPMs value – puts this perfectly in a recent tweet.
What does this mean for managing my ad account?
To begin with, stop paying attention to CPMs. They aren’t actionable.
And neither are your clickthrough rates.
There is only one thing that I use day-in-day-out to really judge performance. That is your cost per action.
Meta optimises towards an action. That action should always be as far down the funnel as you can get away with: purchase/subsription start/trial start/lead. Whatever is the furthest item down the funnel you can get good volume behind.
If Meta is optimising towards that action, it doesn’t matter what your CPM is or your CTR. It also doens’t matter what your thumb stop rate is or your viewthrough (another post on those another time).
Your ad performance will fluctuate. There are days when Meta can find people who want your things, and days when they don’t.
Test creatives based on CPA only. High engagement, good click thoughts. These things are vanity. It is the action you should be focused on.
And anyway you’re a growth marketer or an entrepreneur. Even if all these external factors did have this impact, it’s not your job to let that be the case. It’s your job to buck the trend. You find a way to win when stuff is stacked against you.
As a former boss once said in a meeting when summer was cited as a reason for poor performance, “never come to this meeting with external factors again.” He was right.
Correlation Coefficeints. Applied Statistics, Andrews.