What Does CPM Mean in Advertising?

Your ad account shows a CPM of $14.32 this morning. Yesterday it read $11.80.

Nobody touched the creative. Nobody changed the audience. The number just moved, and now Thursday's budget buys fewer impressions than Monday's did.

CPM stands for cost per mille: what you pay for every 1,000 times an ad loads on a screen, whether or not anyone clicks it. That's the whole definition, and most of the confusion around the term comes from what people expect that number to also tell them.

It doesn't measure quality. It doesn't measure whether the right person saw the ad. It measures one thing: the price of exposure, sold by the thousand.

Everything below is the math behind that price, what counts as reasonable in 2026, and where a flat number beats an auction that moves under you while you sleep.

What CPM Actually Means

Mille is Latin for a thousand, and CPM inherited the word from print. Newspapers and magazines sold ad space by the "thousand circulation" a century before anyone had heard of a banner ad, and when display advertising showed up online in the 1990s it borrowed the unit wholesale, according to Wikipedia's entry on cost per mille.

Ask "what is CPM in advertising" today and the answer hasn't drifted from that original idea. It's a rate: the cost of 1,000 impressions, where an impression is one load of the ad.

But not every platform counts a load the same way. A served impression counts the moment the ad's code fires, whether it's ever on screen. A viewable impression only counts once at least half the ad's pixels sit on screen for a full second, the standard the Media Rating Council set and that most major platforms now report against separately.

That gap matters more than the CPM number itself. Google Display Network's open exchange often clears at 50 to 70% viewability, while full-screen mobile formats on Meta and TikTok clear above 95%, per Prooflytics' 2026 platform benchmark notes. Two placements quoting the same $5 CPM can deliver very different numbers of impressions a human actually saw.

CPM is also a delivery metric, not a performance one. A $4 CPM nobody notices and a $4 CPM that stops the scroll cost the same and count the same, because the meter only counts loads.

Two things follow from that. CPM is the right lens when the goal is reach: getting a message in front of a set number of eyeballs, the way a launch teaser or a brand refresh needs. It's the wrong lens alone for anything meant to convert, because a cheap CPM against a dead audience is a worse buy than an expensive one that actually lands.

The CPM Formula, Worked With Real Numbers

The formula has three steps, and it never changes across platforms:

1. Take the total amount spent on the placement.
2. Divide it by the total number of impressions delivered.
3. Multiply the result by 1,000.

Written out: CPM = (cost ÷ impressions) × 1,000, a version confirmed by Adjust's glossary entry on cost per mille.

Run it with real numbers. A campaign spends $2,000 and delivers 500,000 impressions.

2,000 ÷ 500,000 = 0.004. Multiply by 1,000 and the CPM is $4.00.

Two other versions of that formula are worth keeping on hand, because a media rep will quote you one direction and you'll need to check the other.

To find the cost for a target number of impressions at a known CPM: cost = (CPM × impressions) ÷ 1,000. A $6 CPM buying 250,000 impressions costs $1,500.

To find how many impressions a fixed budget buys: impressions = (budget ÷ CPM) × 1,000. A $500 budget at an $8 CPM buys 62,500 impressions, not 62,500,000, which is the arithmetic slip that turns a Tuesday budget meeting into a Wednesday apology.

Before signing off on a quote, run it backward in three steps: ask for the raw cost and the raw impression count separately, not just the CPM the rep states; divide them yourself with the formula above; and ask whether the impression count is served or viewable, because a "$5 CPM" quoted on served impressions can be a $9 or $10 CPM once you filter to what a person actually saw.

The most common inflation trick isn't a fake CPM at all. It's an inflated impression count that includes bot traffic or unviewable placements, which makes a real $9 CPM look like a $5 one on paper.

Run your own numbers on The Board's CPM calculator before you take anyone else's math on faith, including this post's. Plug in what a platform quoted you and what it actually delivered, and the gap between the two tells you more than the quote ever did.

What Counts as a Good CPM, by Platform

There's no single "good" CPM, because the number is a function of how competitive the audience is, not how well the campaign is run. A CPM comparison only means something inside the same platform and the same objective.

Ranges from Prooflytics' 2026 benchmarks, cited above.

The gap between Google Display and LinkedIn isn't LinkedIn being twenty times worse. It's twenty times more precise about who sees the ad.

B2B buyers pay for that precision because the audience is small and specific rather than broad and cheap.

A separate 2026 report on B2B ad spend makes the same point with a different unit.

Across 153 B2B advertisers and $57.6 million of 2025 spend, tracked by Metadata.io, LinkedIn's average CPM ran $63.19 against Google Ads' $617.91.

That looks alarming until you notice Google's figure comes from search impressions bundled with a 6.33% click-through rate and a $9.76 cost per click. Comparing CPM straight across a feed platform and a search platform is comparing units that were never the same thing to begin with.

This week's action: pull your account's CPM for the last 30 days on one platform only, and check it against the range above for that platform and objective. If you're more than 20% over the range for a comparable audience size, the fix almost never starts with the bid. It starts with frequency, which is the next section's problem.

CPM alone also can't tell you if the traffic it buys is worth having; that's a click-through rate question, answered by what share of those impressions actually turned into a click.

CPM vs CPC: Which One Your Campaign Actually Needs

CPC is cost per click: you pay only when someone acts, not every time the ad loads. Criteo's comparison of CPC and CPM puts the split cleanly: CPM buys exposure regardless of outcome, CPC buys an action and nothing else.

Neither model is inherently cheaper. They convert into each other through click-through rate, and that's the number deciding which one actually wins for a given campaign.

The conversion runs like this: effective CPC = CPM ÷ (10 × CTR expressed as a percentage). A $10 CPM at a 0.5% CTR gives an effective CPC of $10 ÷ (10 × 0.5) = $2.00.

Drop that same $10 CPM to a 0.2% CTR and the effective CPC jumps to $5.00, because the same thousand impressions now produce two clicks instead of five. The CPM never moved. The economics did.

Push CTR up to 1.2% instead and the effective CPC falls to $0.83, cheaper than most direct CPC buys on the same platform. A CPM buy with strong creative can quietly outperform a CPC buy with weak creative, at the identical spend.

That's the actual decision rule: run a CPM buy when the point is impressions themselves, awareness, launch reach, retargeting frequency. Switch to CPC the moment the point is an action, because CPC caps your downside at the click and CPM doesn't cap it at all if the creative goes stale.

A practical test for this week: take your current CPM and current CTR, run the effective-CPC formula above, and compare the result to what a CPC buy on the same platform is actually quoting. If your effective CPC beats the CPC quote by a wide margin, stay on CPM. If it's close or worse, switch, because you're paying the CPM tax without the CPM benefit of scale.

Once a click lands, translating it into the cost of a website visitor is the next number worth checking, since a click and a landed visit aren't always the same thing once redirects and bounces are counted.

eCPM: The Number That Tells You If the CPM Was Worth It

eCPM, effective cost per mille, folds an actual outcome back into the thousand-impression unit instead of just counting spend. Adjust's glossary frames it as the revenue side of the same formula: (total revenue ÷ impressions) × 1,000, instead of cost.

Publishers use it to rank ad units sold under completely different pricing models. A CPC unit and a CPM unit don't compare on their face, so mediation platforms convert every unit back to its effective CPM and rank by that number.

Here's the worked version. Network A pays a flat $6.00 CPM. Network B pays $2.00 per click on the same inventory, at an average 0.4% click-through rate.

Convert Network B to eCPM: 0.004 × 1,000 = 4 clicks per thousand impressions, times $2.00, equals $8.00 per thousand. Network B beats Network A's flat $6.00 rate outright, even though its headline number, $2.00, looked far smaller on the rate card.

The advertiser-side version of the same trap: a $3 CPM looks cheaper than a $9 CPM on a report, and it usually is, until the $3 inventory sits below the fold at 40% viewability and the $9 inventory sits above it at 95%. Fold viewability and conversion back into the number and the $9 CPM often produces a lower cost per outcome.

The action: before comparing two placements on CPM alone, multiply each by its viewability rate and its own conversion rate, and rank on that adjusted number instead of the sticker price. Once a click becomes a sale, the ROAS formula is the next math worth running, because eCPM only gets you as far as the click.

Why CPM Keeps Rising (and the Three Levers That Bring It Down)

A CPM that climbs on its own, with no changes to targeting or creative, has one of three causes almost every time.

Frequency fatigue is the first and most common. The same audience seeing the same creative more than three or four times in a week gets less responsive, and platforms price that drop in engagement straight into a higher CPM.

The fix: pull your frequency metric for the last 14 days. Past 3.5 in a week on a static creative, swap the creative before you touch the budget.

A dropping relevance score is the second. Every major platform scores an ad against how the audience responds to it, and a falling score raises what the platform charges to keep delivering it.

The fix: check click-through rate against the campaign's own opening week. A CTR that's fallen by a third or more while spend stayed flat is a relevance problem, not a market problem, and a new headline or thumbnail usually resets it faster than a bid change does.

Seasonal auction pressure is the third, and it's the one nobody controls. Every advertiser in a category raises budget in the same weeks (Q4 retail, back-to-school, tax season for finance brands), and CPM rises across the whole platform because supply is fixed and demand isn't.

The fix here isn't a lever on your account at all: it's timing the campaign a week earlier or later than the crowd, or accepting the seasonal premium because the audience is worth it that specific week. Sometimes the honest answer is that the cheapest way to advertise that week is a different channel entirely, not a lower bid on the same one.

None of the three get fixed by raising the bid. Raising the bid buys back lost delivery on a campaign that's already fatigued or already irrelevant. It doesn't fix what made the CPM climb, and it usually just moves the same problem to a higher price.

If your CPM is climbing at the same time your customer acquisition cost is climbing, that's the frequency-fatigue pattern showing up twice on the same account, and the creative refresh above fixes both numbers at once.

Where a Flat Price Beats the Impression Auction

Every platform above runs an auction, which is exactly why the CPM at 9am and the CPM at 9pm on the same audience can differ by a third: more advertisers bidding into the same inventory raises the clearing price in real time, and nobody involved gets a vote on it.

The Board skips the auction entirely. It's a public leaderboard where a brand pays a fixed, disclosed amount for a rank, sorted purely by what's been paid.

Reach on the board follows rank instead of a real-time bid: a higher rank delivers more impressions, and that relationship is public on every listing, but the price paid for that rank never moves once it's paid. No bidding war at 9pm, no frequency penalty, no relevance score deciding what you owe tomorrow for a placement bought today.

Taking the top spot right now costs $201, a single number instead of a moving one, and every listing on the board discloses exactly what it paid and what impressions and clicks it's getting in return.

That's not an argument that CPM is broken. It's a genuinely useful unit for comparing display inventory, and this post exists because the math behind it is worth knowing cold.

It's an argument that "impressions, sold by auction" isn't the only model on the table, and a founder deciding where next month's marketing dollar goes should know both exist before picking one.

Pull your last 30 days of CPM data this afternoon, run it against the platform ranges above, and check your frequency metric before you touch a single bid. The rest of what's tracked under Traffic walks through the other numbers worth checking once that one's under control.