Clickthrough rates

Danger.

We’ve recently begun work on our ‘2025 in Paid Social’ report, which will be launching in November. The desired outcome is a study of different ad structures, accounts, and defining what we see as best practice built on all of our data.

One aspect of that is wanting to put some data behind things that often have assumptions to them. Things like thumbstop rate. Or CPM.

But the big one we’re going to talk about this year is CTR.

This is a bit of a preview of some of that work.

CTRs are one of the most useless metrics in paid social

Let’s get the tl;dr out of the way.

Stop using clickthrough rates as part of your decision-making on Facebook right now.

Clickthrough rates do not correlate with CPA.

That means that just because a CTR is higher, does not mean the CPA will be lower.

This statement seems to infuriate some people. If you’ve never had this conversation before or looked at this data yourself, you too right now may be thinking: ‘what? of course they correlate!’ But the data on this doesn’t lie.

Here’s a couple of charts from one dataset I’ve been reviewing recently.

The lefthand side uses all clicks data and the right only unique outbound clicks.

Neither has a statistical correlation.

• All clicks CTR and CPA, correlation coefficient: 0.05

• Unique outbound CTR and CPA, correlation coefficient: -0.04

With correlation coefficients, the closer to +/- 1, the stronger your correlation. The closer to 0, the closer to no correlation.

These two charts are pulled from one Meta account that spends between £50-£150k/month.

What this means is a low, moderate, or high click through rate may well have a low, moderate, or high CPA.

Better CTR doesn’t mean anything in purchase terms.

It’s not just our own data either.

When I posted about this recently on LinkedIn I had more positive feedback from Meta employees than on any other single post I’ve made. Why? I messaged a dozen or so of those who interacted positively with it and the response was uniform: “you’re preaching to the converted, we see this all the time.”

It isn’t a new discovery either.

Back in 2018, I was at Meta HQ in a disruptors/fast growing/spenders of tomorrow cohort, when we were shown this slide.

Clicks don’t correlate to ad recall, brand awareness, or importantly purchase intent.

So why is this happening then? How could it be possible that a click – a necessary step in the journey to become a customer – doesn’t correlate with purchase?

To answer that we have to consider how the Meta auction works.

Every time you open Instagram and have a browse and the Meta algo is preparing to serve you an ad, it makes this calculation.

We also have to remember that the Meta platform, like any computer software, is highly prescriptive.

Anyone who has ever done a class of computer science knows how specific you need to be in giving a computer instructions.

When we run sales conversions campaign with the performance goal of the maximum number of conversions, that is exactly what Meta will do.

It’s agnostic as to who that person is. All it cares about is getting you the conversion – your desired result.

You could have 1,000 reach and 50 conversions, and 10,000 reach and 50 conversions and there’s in essence no difference between those two things.

Similarly, you could have 1,000 reach, 250 outbound clicks, and 50 conversions.

Or you could have 1,000 reach 75 outbound clicks, and 50 conversions.

The outcome of those things is the same. All we’ve told it to do is get us conversions.

So where do clicks come into it?

Well other things to weigh up are the words “estimated” in action rates and in the user value weighting. Estimated signals to us that it predicts and models much of this behaviour. And user value could include and cover a lot.

Why 50 conversion events per week is so important

50 conversion events per week is the joyous moment when theres enough data for the auction to work efficiently.

Below that number, and it is using other proxy metrics to predict expected action rate and user value.

At this stage it takes into account:

• Engagement rates on the ads

• Clicks to the site

• On-site behaviour like bounce rate and dwell time

It does not include mid-funnel metrics like add to cart or initiate checkout.

And so when you’re below 50 conversion events per week, you should do everything humanly possible to get to that number. If that means mass consolidation, then do it. If that means focusing on only one product area, then do it. If it means reducing your account down to a single ad set, then do it. Do whatever you can to get to the 50.

Because most of this data doesn’t correlate to purchase. And so if you’re in this area, you’re in a no mans land of data and performance.

If for whatever reason you can’t get to 50. Then you’re going to need to find a truly correlative higher funnel event that will allow you the chance to reach it.

If you focus on CTR, you may also be ignoring Gen Z

A final thought.

Media measurement platform Measured recently published a study on how Gen Z don’t click. Hat tip on this to Steven Johnson.

In it, they found that Gen Z more than any other generation are more likely to purchase from a view conversion over a click.

We’re working on this data now across our Facebook data, and there’s proving to be some quite interesting aspects to it. Watch this space.