Updated June 20, 2026.
This brand was spending about $2 million a month on Meta ads. At that scale, a creative bottleneck becomes a finance problem. A 54% lift in ROAS changes the revenue profile of the account. A 28% CPM reduction means the same budget buys about 39% more impressions than it did before.
They shared their figures with us for this case study. The brand name, category, and product details stay anonymized.
The account had a problem most large DTC operators will recognize. CPMs kept climbing, Meta Advantage+ had absorbed more of the targeting work, and the one input that still moved the account was fresh creative. The team knew that. They just couldn't make enough of it.
Every winning ad burned out in a couple of weeks. New angles took too long to find. By the time the team shipped the next batch, spend had already drifted into weaker pockets of traffic.
So they handed their Meta and Google ads to Hyper and let the agent run the loop.
What the test measured
This case study compares the brand's Hyper-run period against its prior baseline. The core measures were ROAS, CPM, creative testing speed, and the operating cadence behind the account.
The clean read is that Hyper changed the account in 3 places:
- It increased the number of creative tests the team could run.
- It improved the quality of the creative briefs going into those tests.
- It used segmentation and account data to keep broad Meta delivery from drifting into lower-intent traffic.
The 54% ROAS lift is the outcome. The 20x testing speed and 28% CPM reduction explain how the account got there.
What changed
The fix wasn't a clever setting. It was throughput and judgment, applied every day.
Constant cross-platform creative intelligence. Hyper scanned Instagram and TikTok for what was trending in the category, Reddit for what customers were actually saying, and the Meta Ad Library for competitor ads running in market. Each signal turned into a creative brief with the angle, references, and start frames the team could generate from.
A testing loop that kept moving. The agent watched for fatigue, launched fresh variations, paused weak ads, and scaled winners. The team went from a handful of new ads per week to a pipeline that kept the account supplied.
Segmentation against broad delivery. Advantage+ is powerful, and Meta's Andromeda retrieval system changed how much of the ad-selection work happens inside the platform. Hyper used customer segments, performance history, and conversion signals to shape where spend went.
Execution inside the account. The important part is that the recommendations shipped. Hyper didn't hand the team another report. It built, launched, read, and changed the campaigns.
Inside the creative engine
The creative engine ran like a daily newsroom for paid ads.
Each morning, the agent checked performance by campaign, creative, product, audience segment, and landing page. Then it looked outside the account. Instagram and TikTok showed what the category was starting to copy. Reddit showed the language customers used when they were irritated, excited, or confused. The Meta Ad Library showed where competitors were putting budget.
That outside scan mattered because the brand's best creative ideas rarely came from staring at the ad account. They came from the market.
The agent turned the scan into briefs the creative team could use immediately:
- The customer problem to dramatize.
- The exact phrase or objection that made the angle worth testing.
- The competitor frame to answer or avoid.
- The start frame, offer, product shot, and hook.
- The first 3 variations to generate.
New creative launched into broad targeting in small isolated tests, so the agent could read which ad moved the number. Fatiguing ads were swapped before they dragged down the account. Winners got budget. Losers got retired without a meeting.
| Operating metric | Before Hyper | With Hyper | Why it mattered |
|---|---|---|---|
| Creative test speed | Manual production cadence | Roughly 20x faster | More shots on goal before winners fatigued |
| Trend discovery | Competitor review and internal brainstorming | Instagram, TikTok, Reddit, Meta Ad Library, and account data | Briefs started from live market signals |
| Delivery control | Mostly platform automation | Platform automation plus customer segmentation | Less spend leaked into low-intent traffic |
| Decision latency | Human review between insight and launch | Agent executed the change | Good ideas reached the account faster |
The numbers
Over 12 months:
- ROAS came up about 54% versus the brand's prior baseline.
- CPMs came down about 28%.
- Creative was tested roughly 20x faster.
- A 28% CPM reduction means the same media budget bought about 39% more impressions, before accounting for the ROAS lift.
- On $2M/month in Meta spend, that combination changed both efficiency and scale.
| Metric | Result | How to read it |
|---|---|---|
| ROAS | +54% | For every 1.00 USD of prior return, the Hyper-run period returned about 1.54 USD |
| CPM | -28% | The same budget bought about 39% more impressions at the new CPM |
| Creative testing speed | 20x faster | The account found winners sooner and replaced tired ads sooner |
| Spend scale | $2M/month | Efficiency changes mattered because the account was already spending at scale |
| Primary driver | Creative throughput plus segmentation | More useful creative entered the auction, with cleaner buyer signals |
The mechanism was simple in hindsight. More tests meant more chances to find winners. Better briefs made each batch more likely to contain something useful. Better segmentation reduced wasted delivery. The brand stopped losing the race against creative fatigue because it was no longer running that race by hand.
What the numbers mean on a $2M account
Percentages can hide the size of the swing.
At $2M/month in media spend, a 28% CPM reduction is a major supply gain. If the brand previously bought 100 impressions for a given budget, the same spend bought about 139 impressions after the CPM decrease. The extra impressions still need to convert. They give the account more qualified inventory to test against.
The ROAS lift is the cleaner business read. A 54% ROAS increase means the account produced 1.54x the return of its prior baseline. The customer did not share enough margin data for us to publish contribution-profit numbers, so we won't pretend we have them. The point is still clear: when spend is already $2M/month, a better testing loop changes the economics quickly.
An unexpected finding
The best-performing creative angle of the quarter didn't come from the Meta Ad Library. It came from Reddit.
The agent caught a recurring complaint in the brand's subreddit, weeks before any competitor turned it into an ad. Customers were describing a problem in almost the same words, over and over. The product solved that problem, and the brand had never made it the center of a paid concept.
The creative built around that phrasing became the brand's top performer.
That finding is the best argument for why AI ad agents need to watch more than ad platforms. Competitor ads show what the market has already seen. Customer language shows what the market is ready to hear next.
How this case study fits with the other Hyper tests
This is one of the accounts behind the numbers in how we built Hyper, the best AI for Meta and Google ads.
For the head-to-head version, where an agency ran half its accounts on Hyper and half by hand, read the agency benchmark case study. For the operating-labor side, read the original case study, where one team freed up 29 hours a week.
Together, the 3 stories show the same loop from different angles:
- The ecommerce account shows performance at spend scale.
- The agency benchmark shows AI versus a human team on comparable client accounts.
- The original case study shows how much time disappears when reporting and campaign setup stop being manual work.
How Hyper helps
Hyper is an AI marketing agent for operators who need Meta and Google ads to improve without adding another weekly dashboard ritual. It watches the account, reads the market, builds creative briefs, launches tests, trims waste, and writes the performance readout.
For ecommerce teams, the useful part is the loop: research, creative, launch, optimize, report, repeat. The same agent that finds the angle can ship the test and read the result.
Autonomous marketing
Grow your business faster with AI agents
- Automates Google, Meta + 5 more platforms
- Handles your SEO end to end
- Improves website conversions
- Runs social media for you
Frequently Asked Questions
Frequently asked questions
Q: What did Hyper improve in this ecommerce ads case study?
Hyper lifted ROAS by about 54%, lowered CPMs by about 28%, and increased creative testing speed by roughly 20x over 12 months. The customer was spending about $2M/month on Meta ads, so the efficiency gains mattered at real scale.
Q: How did Hyper improve ROAS?
The lift came from faster creative testing, better creative briefs from market signals, and cleaner customer segmentation. Hyper scanned Instagram, TikTok, Reddit, the Meta Ad Library, and account data, then launched and optimized tests inside Meta and Google.
Q: Why does CPM matter in an AI ads case study?
CPM tells you how much inventory the same budget can buy. A 28% CPM reduction means the same spend buys about 39% more impressions. For a $2M/month account, that gives the testing loop more room to find profitable creative.
Q: Is this case study anonymized?
Yes. The customer shared the performance figures for this case study. The brand, category, and product details are anonymized.