Meta is still the default channel for most of the e-commerce and DTC brands we work with, but it’s gotten weird in 2026. Touted as the agency killer, Meta’s changes are more of a dumb robot traffic jam. That’s what happens when everyone takes the same routes to success.
Thank you for reading this post, don't forget to subscribe!The Benchmark Numbers
Rough ranges for ecommerce advertisers in 2026:
| Metric | Typical Range |
|---|---|
| CPM (cost per 1,000 impressions) | $11–$14, often near $14 |
| CPC (cost per click) | $0.75–$1.05 |
| CPA (cost per purchase) | $32–$53, depending on vertical |
| Minimum daily budget to test | $21–$52/day |
| Monthly budget to properly fund 1–2 ad sets | $1,600–$3,200/mo |
One caveat. Lead-gen costs and ecommerce purchase costs get blended together in a lot of industry reports, and they shouldn’t be, a click on a lead form and a purchase are different economics. These are the ecommerce numbers, since that’s what applies to most of our clients. Your category matters more than anything else on this list.
What’s Actually Driving Costs Up
Costs across Meta climbed roughly 20% year over year. Easy read: AI automation isn’t working. Wrong read. Here’s what’s actually happening.
The efficiency gains are getting competed away. Everyone in your category is running the same automated bidding tools now. Individually, the automation works fine, better buyers, faster creative testing, real-time bid adjustments. Collectively, it means every advertiser got sharper at the same time, chasing the same impressions. That bids the price up. The AI didn’t fail. It worked for everyone at once, and that’s worse in aggregate than it is for any single account. Automation stops being an edge the moment it’s what everyone’s running.
The AI is only as good as what it can see. Most advertisers miss this. If Meta’s algorithm only receives a fraction of your actual conversions, because of iOS restrictions, consent banners, or checkout flows that don’t fire the pixel cleanly, it’s optimizing against a broken picture of who your buyers are. That’s not the algorithm being dumb. That’s the algorithm getting bad inputs and still being expected to perform. Half of “AI isn’t working” is really “we never gave it complete data.”
Your vertical still sets the floor. Lifestyle and boutique runs closer to $32 CPA, electronics pushes toward $53. Automation doesn’t change that. Your industry sets your baseline before you’ve touched anything.
Seasonality still stacks on top of all of it. CPMs climb from October, peak in November, spike hardest on Black Friday and Cyber Monday, commonly two to three times the annual baseline.
And there’s a platform-incentive angle worth saying plainly. Meta benefits when automation stays on by default and is hard to fully kill. More of your spend flows through systems you can’t see into or price-shop against. Be skeptical of “just trust the algorithm,” not because the algorithm is dumb, but because Meta’s incentives and yours don’t fully line up. The fact that you have to fight to keep automation off just to get a clean read is the tell.
Cost Rising Doesn’t Automatically Mean ROI Is Falling
Check this before you conclude automation is hurting your account. If AOV, conversion quality, or LTV are also climbing, a higher CPA can still be a good trade. Rising cost per result is worth watching, but on its own it doesn’t prove automation is net-negative, it proves the marginal impression got more expensive. Look at blended ROAS or profit per dollar, not CPA alone, before you write it off.
Budgeting to Actually Get Data
Meta’s algorithm needs volume to learn, automated or not. Roughly 50 purchase events in a rolling seven-day window per ad set gets you out of the learning phase. Rough formula:
daily budget ≈ (target CPA × 50) ÷ 7
At a $30 target CPA, that’s about $215/day per ad set to clear learning in a week. Spread too thin across too many ad sets and none of them get enough data to stabilize. That’s the most common reason a Meta test never gets a fair shot, automated or manual.
When scaling, move budgets up in steps of around 20%, not overnight doubles. Big jumps and mid-flight edits to audience or optimization event reset the learning phase and put you back at zero.
Meta vs. Google vs. TikTok, Quickly
| Meta | TikTok | ||
|---|---|---|---|
| Average CPM | $11–$14 | Display ~$3, Search is CPC | ~$11.50 |
| Average CPC | $0.75–$1.05 | Ecom median ~$0.45–$0.50 | ~$1.15 |
| Creative lift | Moderate | Lower, product feed driven | High, native video required |
| Audience intent | Discovery and browsing | Active search intent | Discovery and browsing |
Meta and TikTok reach people before they’re looking. Google catches people already searching, narrower audience, lower cost per click. Most of our accounts run Meta as the core prospecting and retargeting engine, layer Google in for the bottom of the funnel, add TikTok once the first two are working.
What We’d Actually Do About It
Treat “let the algorithm handle it” as something to test, not something to trust by default. Two things matter more than whether automation is on or off.
Fix your tracking before you touch your bidding strategy. Server-side conversion tracking next to your pixel closes most of the signal gap that’s quietly wrecking both your reported numbers and the algorithm’s ability to optimize. Half the time, the ads are working better than the dashboard says, because the algorithm’s been starved of good data.
Don’t assume automated and manual bidding land in the same place for your account. Test it on your own data, long enough to clear the learning phase, before you decide either way. Platform-wide costs rising is real. Whether that means automation specifically is failing you is a different question, and only your own numbers answer it.
If you want a second set of eyes on what your account is actually costing versus what it’s reporting, let’s talk.
