What Is a Good CPM?

Sourced benchmarks put Meta's CPM between $8.19 (Gupta Media, Social CPM Tracker, Oct 2025) and $15.06 (Triple Whale, via Fibbler, Sep 2026) per 1,000 impressions, with Gupta Media's TikTok, YouTube and Pinterest averages lower still. Within those ranges, a good CPM is any price under your CPM ceiling: the most you can pay per 1,000 impressions and still win customers at your target cost per acquisition.

So what is a good CPM has a market answer and a personal one. The market answer is a table of dated, linked benchmarks that rarely agree; the personal one is a ceiling you can compute from three numbers in your own ad account.

CPM benchmarks by platform

The table lists the CPM benchmarks we could trace to a readable source page on September 30, 2026. None comes from the ad platforms, because none of the platform help pages we checked publishes an average price. Every figure below is a third-party measurement with its own sample.

Read the table by source, not by row

Within one tracker the platforms line up cleanly. Gupta Media's 2025 averages run from $4.67 on Pinterest to $8.60 on Snapchat, with Meta near the top. Across trackers, the same platform's figures can differ by a factor of 1.8, which the Meta section below unpacks.

X (Twitter) is missing on purpose. The only X CPM we could fetch had no stated basis, so we left it out rather than print a number nobody can check. For TikTok, both trackers land close together, and the TikTok CPM benchmarks sheet breaks that figure down by bidding type. The creative side of running there is on the page about making TikTok video ads with a saved AI actor.

What is a good CPM for Meta ads?

Meta sells Facebook and Instagram placements from the same Ads Manager, and the two sourced reference points are $8.19 (Gupta Media, Social CPM Tracker, Oct 2025) and $15.06 (Triple Whale, via Fibbler, Sep 2026). A Meta CPM between them is ordinary. Below the lower one you are buying cheap reach; above the higher one, check audience size and objective before you assume a problem.

The two numbers differ for reasons you can name. Gupta Media's is a 2025 average across many industries, mostly in the US. Triple Whale's is a median across 40,000+ ecommerce brands only, covering Aug 2025–Jul 2026. We read it through Fibbler's summary because Triple Whale's own page blocked our fetch.

To read a Meta CPM properly you need the click, conversion and acquisition figures downstream of it. These are the sourced ones:

No source we checked publishes an Instagram-only CPM separate from Meta's combined figure. Metadata.io's B2B data does split cost per lead: $138 (Metadata.io, 2026 B2B Paid Media Benchmark Report, 2026) on Instagram against $145 on Facebook, so for B2B lead generation the two placements priced close together.

Creative formats for each placement are covered on the pages for Facebook video ad creative and Instagram Reels and feed ads.

Why CPM benchmarks disagree

Two honest trackers can publish very different CPMs for the same platform. Before you compare your own number with any benchmark, check these things about the source:

• Sample. Ecommerce-only datasets such as Triple Whale's and Lebesgue's measure brands bidding for purchases. Multi-industry trackers such as Gupta Media's also include awareness and traffic buyers.
• Objective. Conversion-optimised delivery buys impressions from people likely to act. TikTok writes this into its definition of oCPM, and those impressions are not the same product as plain reach.
• Statistic. A median ignores a handful of very expensive accounts; an average does not. A spend-weighted figure leans towards the biggest advertisers.
• Region and currency. Most trackers are mostly US or do not say. Dreamdata reports LinkedIn in euros, so its figure cannot be read as dollars.
• Age. Adzoola's YouTube data covers February 2024 to January 2025, more than a year before this page was checked.
• What counts as an impression. Google's viewable CPM counts an impression only when half of a display ad is on screen for one second, or a video ad plays for two seconds. A plain CPM counts every impression served.
• Season. A full-year average and a single month are not comparable, because auction prices move with how many advertisers are bidding at the time.

The practical rule: compare your CPM with the benchmark whose sample, objective and period are closest to your campaign, and treat the others as context.

Is a high CPM good or bad?

For an advertiser, a high CPM is neither good nor bad on its own. It is a price, and what matters is what the impressions return. For a publisher or creator selling ad space, the reverse holds: a high CPM, or eCPM, means more revenue per 1,000 impressions served.

The LinkedIn row shows why an expensive impression can still be a good buy. Dreamdata measured LinkedIn's 2025 B2B CPM at €34.33, far above any other platform in the table.

Yet the same report put LinkedIn's attributed return at 121% (Dreamdata, LinkedIn Ads Benchmarks Report 2026, 2026-03-10), ahead of Google Search at 67% and Meta at 51%. Those are B2B attributed-revenue figures, not ecommerce ROAS, but the pattern holds: expensive reach to the right buyer can beat cheap reach to the wrong one.

A low CPM can equally be a warning. Cheap impressions often come from placements or audiences that rarely click, and the next section shows how a $6 CPM can cost more per customer than a $32 one.

Judge a CPM by the CPA it implies

A CPM becomes a cost per customer through two rates: the share of viewers who click (CTR) and the share of clickers who buy (conversion rate). The relationship is exact, with both rates written as decimals:

The table runs four example campaigns through it. The inputs are round illustrative numbers, not benchmarks.

Campaign A buys the cheapest impressions and the most expensive customers. Campaign D pays 5.3 times as much per 1,000 impressions and still acquires customers for $40 rather than $66.67, because a CTR 3.3 times higher and a much better conversion rate outweigh the price.

Work out your CPM ceiling

Turn the CPM formula around and you get the highest CPM your campaign can afford. Any CPM under that ceiling is a good CPM for you, whatever a benchmark says, and any CPM above it is too expensive, however ordinary it looks in a table.

Worked example: an order worth $120 at a 35% gross margin leaves $42 before advertising, so the break-even CPA is $42. At a 1% CTR and a 2% conversion rate, the ceiling is $42 × 1,000 × 0.01 × 0.02 = $8.40. At a 1.5% CTR and a 3% conversion rate, it rises to $18.90.

So is a $12 CPM good? For the first campaign, no: it implies a $60 CPA against a $42 target. For the second, yes: the same CPM implies $26.67.

The free CPM, CPC, CTR and ROAS calculator runs this chain on your own numbers, and the break-even ROAS and CPA calculator gives you the target CPA to test against. For the revenue side of the same test, see what a good ROAS is against your break-even.

Common reasons a CPM runs above the range

Across platforms, an above-range CPM usually traces back to one of six things:

• A narrow audience. Fewer eligible people means more advertisers competing for each of them.
• A conversion objective. Delivery optimised for sales or sign-ups is priced as a different product from plain reach, and usually costs more per impression.
• Competition peaks. Retail holidays and big launches in your category crowd the auction.
• Low predicted relevance. Auctions weigh the bid and the expected response together. TikTok's bidding help page says its auction "ranks ads based on bid price and the relevance an ad might have to people". Weak engagement therefore shows up as a higher price.
• Fatigue. Showing the same few ads to the same people wears down engagement, and relevance with it.
• Placement and region. Premium placements and high-income markets cost more per impression.

Four of the six are targeting and scheduling choices. Relevance and fatigue are creative problems, and the final section prices the fix.

Creative testing and your CPM ceiling

Creative moves the CPM ceiling, not the auction's price list. Using the example above, a variant that lifts CTR from 1% to 1.5% at the same 2% conversion rate raises the ceiling from $8.40 to $12.60. The only way to find that variant is to test several, and testing is where creative cost bites.

Hiring creators prices every variant separately. At Billo's compiled market average of $200 per video (Billo, UGC Rates, 2026-06-16), 10 creator videos cost $2,000 before usage rights, which Billo says can push the bill to two or three times as much. Add that to media spend and your all-in CPM sits well above the one the ad account reports.

An AI studio swaps that per-video fee for a subscription. Playcut's paid plans run from $9/mo (Hobby, 500 credits) through $29/mo (Pro, 2,000 credits) and $79/mo (Studio, 6,000 credits) to $149 per seat (Agency, 10,000 credits per seat).

A 10-second 720p clip of a saved AI actor speaking your script costs 85 credits on the default Act tier, so Pro's monthly credits cover 23 such clips before any re-rolls, and a discarded take still counts against them.

None of this lowers your CPM by a number we can promise, and we make no such claim. An AI actor is a presenter, not a customer: it can demonstrate the product but never claim to have bought or used it. The rules are in what the FTC allows in AI UGC ads. The AI UGC ad generator shows the workflow for making variants in bulk.