Why I Don't Shut Up About CPMr

· Articles

Why I Don't Shut Up About CPMr

I Keep Banging On About CPMr. Here's Why.

I talk about CPMr a lot. Rising CPMr is one of the earliest signs that something is wrong with an account: before ROAS drops, before CPA climbs, before anyone on the brand side notices anything is off.

I'm going to show you what that actually looks like in a real account, with nine weeks of data, so next time you see it you'll know exactly what you're looking at.

What CPMr Is

CPMr is CPM multiplied by frequency.

CPM tells you how much it costs to serve 1,000 impressions. Frequency tells you how many times the average person in your audience has seen your ads. Multiply them together and you get a single number that represents the true cost of your reach: accounting for the fact that you're increasingly paying to show ads to the same people again.

When CPMr rises, one of two things is happening. Your CPM is going up, your frequency is going up, or both. Either way, the account is becoming less efficient at reaching new people.

That's the thing to watch. Not ROAS. Not CPA. CPMr. It tells you what's coming before the other metrics catch up.

What It Looks Like in Practice

I'm pulling from a real account here. Numbers are real, client is anonymised.

Year-over-year, January to early March:

  • Spend up 59.1% ($259k to $413k)

  • Reach up 12.5% (4.3M to 4.8M)

  • Cost per 1,000 people reached up 41.4% ($60 to $85)

Spending 60% more to reach 12.5% more people. That gap is CPMr doing exactly what it does when an account has a problem.

But the cost-per-reach number is the outcome. To understand why it's happening, you need to look at net new reach.

Net New Reach Is the Signal

Net new reach is the number of unique individuals seeing your ads for the first time each week. Not total reach, new reach. People who haven't been in your audience before.

Here's how that percentage trended across the first nine weeks of 2026 vs the same period in prior years:

In 2024 and 2025, the net new percentage dropped mid-quarter and then recovered. By W9 of both years it was back above 38%. In 2026, it has declined consistently every week and sits at 26.3% by W9.

That means nearly three quarters of the budget in W9 is being spent re-serving ads to people who have already seen them.

What Causes This

When you increase spend without increasing creative diversity and maintaining exclusions, the algorithm has to work harder to find new people to show your ads to. At some point it runs out of easy options and starts cycling back through the same audience.

The creative isn't necessarily fatigued in the traditional sense: the ads haven't stopped working. The audience is saturated. There aren't enough new people the algorithm can reach with the current setup at the current spend level.

This is negative scaling. By W9 of 2026, this account reached 1.1M people. In W9 of 2025, with significantly less spend, it reached 1.2M. More money, fewer people. That's the saturation trap.

Why CPMr Is the Early Warning

The problem with most of the metrics people watch, ROAS, CPA, revenue, is that they're lagging indicators. By the time they move meaningfully, the issue has been building for weeks.

CPMr moves first. When frequency starts creeping and CPM follows, CPMr is the first number to reflect it. If you're tracking it weekly, you'll see the problem forming before it shows up anywhere else.

In this account, CPMr rising from $60 to $85 year-over-year wasn't a sudden event. It was a slow climb that corresponded directly with the net new reach percentage falling week on week. One caused the other. CPMr was the signal that the reach pool was narrowing.

What to Do When You See It

Rising CPMr tells you the account needs intervention. It doesn't tell you exactly what the intervention should be, that depends on the account, but the order of operations is usually:

  1. Check exclusions. Are you excluding existing customers, recent purchasers, recent website visitors? If not, start there. You're paying to reach people who are already in your funnel.

  2. Check campaign consolidation. Fragmented ad sets competing for the same audience inflate CPM. Consolidating spend gives the algorithm a broader brief.

  3. Check creative diversity. If every ad in the account is conceptually similar, you're limiting who the algorithm can reach. Genuine creative variety - different formats, different angles, different problem territories - opens the reach pool back up.

Cody Plof shared a real example of this recently. His team dropped iCPA by 26% by tackling CPMr directly. The interventions:

  • Moved spend aggressively from video views objective to conversion objective

  • Increased Partnership Ads spend

  • Added a campaign with view optimisation

  • Scaled AEB investment

  • Improved Reels delivery

  • Patched exclusion leaks

Cody Plofker
@codyplof · Mar 3

We just dropped iCPA on Meta 26% by doing the following:

- Moved spend aggressively from VO to CO

- Increased partnership spend

- Added a campaign with view optimization

- Scaled AEB investment

- Improved reels delivery

- Patched some exclusion leaks

All just to improve CPMr. It's wild how much you have to do just to have a chance of success these days.

I'd like to see Manus try.

↗ View on X

His summary: "All just to improve CPMr. It's wild how much you have to do just to have a chance of success these days."

When I asked if CPMr had improved alongside the iCPA drop, the answer was: "Big time."

The Attribution Trap That Makes It Worse

There's a second layer to this that's worth understanding, because it explains why so many accounts end up in the CPMr trap in the first place.

Over-optimising for click attribution pushes the algorithm toward lower-funnel, higher-intent audiences. Those audiences are smaller. When you concentrate spend on them, frequency climbs faster, CPMr rises faster, and the reach pool narrows faster than it would if you were running a broader media mix.

Cody's test made this explicit. When his team made changes to improve CPMr - adding view optimisation, scaling Reels, increasing Partnership Ads - Meta ROAS went down. Last-click performance went down. CTR went down. The in-platform numbers looked worse.

Cody Plofker
@codyplof · Mar 3

We just dropped iCPA on Meta 26% by doing the following:

- Moved spend aggressively from VO to CO

- Increased partnership spend

- Added a campaign with view optimization

- Scaled AEB investment

- Improved reels delivery

- Patched some exclusion leaks

All just to improve…

↗ View on X

Business performance improved.

His point, and one I agree with: incrementality doesn't equal attribution. Not every ad needs to drive a last-click conversion to be doing valuable work. Views can be highly incremental. Reels behave differently to standard feed placements. Running reach or mid-funnel activity with poor reported ROAS can still be the right call if it's expanding the reach pool and keeping CPMr in check.

Leadership and finance teams need to understand this, because the people closest to the data often know what needs to happen but don't have the cover to make changes that will make in-platform numbers look worse in the short term.

The goal isn't a good-looking dashboard. It's a healthy account that can keep reaching new people efficiently over time. CPMr is how you measure whether you're achieving that.

Watch it before the other numbers tell you to.

TL;DR

  • CPMr = CPM x frequency. It rises when the algorithm is re-serving ads to the same people

  • It's a leading indicator - it moves before ROAS and CPA reflect the problem

  • When net new reach % falls, CPMr rises. They move together

  • This account spent 59% more year-over-year to reach 12.5% more people - a direct result of rising CPMr and declining net new reach

  • By W9, 74% of budget was being spent on people who had already seen the ads

  • Fixing it requires exclusions, consolidation, creative diversity, and often a broader media mix including view optimisation and Partnership Ads

  • Over-optimising for click attribution is part of what causes the problem - it concentrates spend on smaller, lower-funnel audiences and accelerates frequency

  • Improving CPMr may make your in-platform numbers look worse in the short term. That doesn't mean it's the wrong call

  • Track CPMr weekly. If it's consistently climbing, the account needs intervention before the headline metrics catch up


Book a call with me.


Back to all articles