Ad costs are up sharply this year — industry analyses put average CPM increases at roughly 20% year over year, with some portfolio studies reporting closer to 28% and individual verticals swinging anywhere from 15% to 40%. Those numbers have prompted a lot of budget panic and a lot of unhelpful advice. The problem is that rising CPMs in 2026 aren't one phenomenon; they're three distinct things wearing the same label, and each demands a completely different response. Diagnose the wrong one and you'll spend months solving a problem you don't have while the actual cause keeps compounding.
Three causes, not one
1. Structural auction pressure. More advertisers competing for broadly fixed inventory. Real, permanent, and not something you can fix — only adapt to.
2. An algorithmic penalty. Meta's Andromeda system reportedly delivers repetitive creative at higher cost. This portion is a penalty you're incurring, not a market price — and it's recoverable.
3. A measurement artefact. Attribution and counting changes made costs appear to rise without underlying performance changing. Nothing broke; the maths did.
Most advertisers are experiencing some blend of all three. The mix determines what you should do.
Cause one: the structural squeeze
The unglamorous truth first. More advertisers keep entering the same auctions, chasing an audience that isn't growing at the same rate. That's textbook price inflation, and it's the portion of your CPM increase that isn't going away. Reported average CPMs in the region of $13 reflect a market where cheap attention has simply stopped existing.
There's no clever fix, only a strategic one: if the price of attention rises permanently, the only durable response is to extract more value from each unit of it. That means conversion rate, average order value, and repeat purchase matter more than they did when traffic was cheap enough to waste. It also means the flat-budget reality we cover in where marketing budgets are shifting bites harder in paid social than almost anywhere else.
Cause two: the Andromeda penalty (the fixable one)
This is where most of the recoverable ground sits, and it's the part generic "CPMs are up" articles miss entirely.
Meta rebuilt its ad retrieval engine — the system rolled out globally by late 2025 and applied to every advertiser by early 2026. The company has described it as a next-generation personalised retrieval system running on specialised hardware, with model capacity increased by orders of magnitude. The practical consequence is a genuine inversion of how targeting works: rather than you selecting audiences, the system reads your creative and decides who should see it.
The inversion Your creative is now your targeting. Which means a shallow creative library isn't just a content problem — it's the reason your delivery costs more.
The mechanism that raises costs is creative similarity. Reporting on the system indicates that accounts running visually repetitive creative — ten ads that are essentially one image with different text overlays — are treated as fatiguing and delivered at higher CPMs. Cosmetic variation doesn't register as diversity. The system wants genuinely different formats, angles, and lengths.
Two consequences follow. Creative fatigue now reportedly sets in within two to three weeks rather than four or more, because a faster retrieval system burns through audiences quicker. And accounts need substantially more creative production volume than they did — roughly double, by several practitioner accounts — simply to stand still. That reframes creative from a design task into an operational one, which is why systematic creative testing and genuine production capacity — the domain of design and creative teams — have become the primary performance lever in paid social.
Atria
AI ad platform that researches, creates, launches, and analyzes winning Meta and TikTok creative.
Best for: Performance marketers scaling Meta and TikTok ads
Cause three: the measurement artefact
The third cause is the one that produces the most unnecessary panic. Attribution and counting changes during 2026 made reported costs per acquisition appear meaningfully higher — figures in the range of 15% to 30% have been reported — without underlying performance actually deteriorating. Some of the "cost crisis" is a reporting crisis.
This matters because the response to a measurement shift is nothing at all, whereas the response to a genuine cost increase is significant work. Advertisers who couldn't tell them apart spent the spring rebuilding campaigns that were performing fine. Distinguishing the two requires measurement you trust — which is exactly why conversion tracking you can verify and honest attribution thinking stopped being back-office concerns this year.
Diagnose before you react
| What you're seeing | Likely cause | Response |
|---|---|---|
| CPMs up, but conversion volume and revenue steady | Measurement change | Verify against backend revenue; change nothing |
| CPMs climbing steadily as creative ages | Creative fatigue penalty | Refresh with genuinely different creative |
| Many ads live but all visually similar | Creative similarity penalty | Vary format, angle and length — not just copy |
| Costs up across every account and vertical you see | Structural inflation | Improve conversion rate and order value |
| Erratic delivery, poor event matching | Weak signal quality | Server-side events, deduplication, one clear objective |
What advertisers are actually doing about it
Producing more, and more varied, creative
The single most common adaptation. Not more variations of one idea — genuinely different concepts, formats and lengths, refreshed on a schedule rather than when performance already dropped. Practitioner guidance clusters around weekly refresh for high-spend accounts and monthly for smaller ones. Catalogue and dynamic product formats have become popular partly because they generate creative diversity automatically, without waiting on production cycles.
Fixing signal quality
Poor conversion data doesn't just distort reporting — it slows the algorithm's learning and inflates delivery costs. Running server-side events alongside the pixel with deduplication enabled, keeping event match quality healthy, and optimising toward one consistent objective rather than switching between them are now baseline hygiene rather than advanced practice.
Consolidating rather than fragmenting
Splitting budget across many small campaigns starves a system that needs volume to learn. Consolidation gives it room, which is the same structural logic we set out in structuring an account for scale and control — different platform, identical principle.
Following the cheap inventory
Every new placement follows the same arc: early adopters get inexpensive inventory, competition arrives, prices normalise. Advertisers are exploiting that window deliberately — vertical short-form placements reportedly carry meaningfully lower CPMs than feed placements, and newer surfaces have been reported at a fraction of established feed costs. It's a genuine arbitrage, but a temporary one by definition — the window closes as soon as enough advertisers notice it.
Spreading beyond the duopoly
As costs climb on the largest platforms, more advertisers are treating Reddit, Pinterest, Snapchat, retail media and connected TV as genuine parts of the mix rather than experiments — not because they replace Meta, but because concentrating an entire budget in the most competitive auction on earth is a strategy with obvious downside.
The reframe that matters most
Here's the shift that separates advertisers who are coping from those who are panicking. CPM is an input price, not an outcome. It tells you what a thousand impressions cost — nothing about whether that purchase was wise.
Paying 30% more per thousand impressions is only a problem if you're doing the same thing with them. If your conversion rate improves by half, your effective cost per customer falls even as CPM rises. That's why the most effective responses this year have often had nothing to do with media buying: better landing pages, stronger offers, and higher-value repeat purchase all reduce your dependence on cheap impressions. It's also why retargeting efficiency deserves scrutiny — in a high-CPM market, paying repeatedly to reach people who would have returned anyway is an expensive habit.
The broader context is that this is what the automated era looks like: the levers that used to control cost have moved, as we discuss in what's left to control in PPC. You no longer bid your way to efficiency. You earn it through creative, signal quality, and what happens after the click.
The bottom line
CPMs are genuinely up in 2026, somewhere in the region of 20% to 30% depending on whose data you read and which vertical you're in. But before you rebuild anything, work out which of the three causes is actually hitting you: structural inflation you can only adapt to, a creative-diversity penalty you can substantially reverse, or a measurement change that requires no action at all. Then act accordingly — more and genuinely varied creative on a faster refresh cycle, clean server-side signal, consolidated campaigns, and deliberate testing of cheaper placements while the window lasts.
One practical note on timing: benchmark against the same period last year rather than last month. CPMs are strongly seasonal, climbing as retail budgets flood the auction in the fourth quarter and easing afterwards, so a month-on-month comparison will regularly tell you a story about the calendar rather than your account. Comparing like periods separates the seasonal swing from the structural trend — and stops you rebuilding a campaign in November because it costs more than it did in September.
And keep the reframe in view: CPM is a price, not a verdict. The advertisers doing well this year aren't the ones who found cheap impressions. They're the ones who got more out of expensive ones.
Ad costs climbing faster than your results?
We diagnose what's actually driving your CPMs — then fix the creative, signal, and conversion levers that bring your cost per customer back down.
Explore Performance Marketing →Frequently asked questions
Why did CPMs rise so much in 2026?
Three separate forces stacked on top of each other. First, structural auction pressure — more advertisers competing for essentially fixed inventory, which is permanent. Second, an algorithmic effect: Meta's Andromeda retrieval system reportedly raises delivery costs for accounts whose creative lacks genuine diversity, meaning part of the increase is a penalty rather than market inflation. Third, attribution and measurement changes made costs appear worse than they became. The responses differ for each, so diagnosing which applies to you matters more than the headline percentage.
How much have Meta CPMs increased in 2026?
Reported figures vary by source and vertical. Industry analyses through early 2026 put average CPM increases at roughly 20% year over year, with one portfolio study across 170+ brands reporting around 28% in the first quarter, and some agencies citing swings of 15% to 40% depending on category. Average CPMs in the region of $13 have been reported. Treat any single figure as directional — your own vertical and account history are the more meaningful benchmark.
What is Meta's Andromeda algorithm?
Andromeda is Meta's rebuilt ad retrieval engine, rolled out globally by late 2025 and applying to all advertisers by early 2026. Meta has described it as a next-generation personalised retrieval system running on specialised hardware with vastly increased model capacity. Practically, it shifts targeting from advertiser-selected audiences toward the system reading your creative to decide who should see it — which is why creative diversity, rather than audience configuration, has become the primary performance lever.
How do you lower CPMs?
The most effective lever is genuine creative diversity — different formats, angles, and lengths rather than cosmetic variations of the same asset, since systems that detect repetitive creative reportedly deliver it at higher cost. Beyond that: refresh creative faster, as fatigue now sets in within a few weeks; improve conversion signal quality with server-side events; use vertical formats suited to where most inventory now sits; consolidate campaigns so the algorithm has room to learn; and test cheaper emerging placements before competition prices them up.
Are rising CPMs permanent?
The structural component almost certainly is. More advertisers competing for a broadly fixed supply of attention pushes prices up over time, and nothing on the horizon reverses that. However, the portion caused by creative fatigue penalties is recoverable through better creative practice, and the portion caused by measurement changes was never a real cost increase at all. Cheap attention isn't returning, but a meaningful share of most advertisers' reported increase is addressable.
