For two years I believed the same comfortable lie that most small media buyers believe: if a campaign stops working, the fix is somewhere in the settings. Budget. Bid strategy. Audience. Placements. Campaign structure, that dark art people write threads about. I have spent whole weekends in there, dragging sliders with the seriousness of a surgeon, and almost none of it mattered. My accounts were never starved of budget. They were starved of ideas.
Who I am
I'm Marco. I buy ads for a living - not at an agency with a floor of people and a creative department, just me, a laptop, and four e-commerce clients who pay me a retainer to keep their Meta and TikTok accounts breathing. Two skincare brands, one coffee subscription, and one very stubborn company that sells dog beds and believes, sincerely, that its product photography is fine.
The job is not what people think it is. Nobody hires me for my opinion on lookalike percentages. They hire me because the number at the bottom of the dashboard has to keep going in the right direction, and when it does not, I am the one who explains why on a Monday call. I like the work. What I did not like was the slow discovery that the part of the job that actually decides the outcome - the creative - was the one part I had no capacity to do at volume.
The week my best ad died
Every account I run has had one ad like this. You know the one. It came out of nowhere, it beat everything you tested for six months, and quietly it became the business. For the coffee client it was a fourteen-second clip of the founder grinding beans badly in his own kitchen, filmed vertically, with a hook about how he could not afford a real studio. It carried something like half the spend on the account. It was not a campaign; it was a load-bearing wall.
In February it started to wobble. Not a crash - a drift. Costs crept up a little each day, the kind of creep you can argue yourself out of noticing for a week and a half. By the time it was undeniable, the number was ugly, and I had nothing warmed up behind it. Every other ad in the account was a variation of the same ad, which meant the whole account fatigued at roughly the same speed, like a family all catching the same cold.
On the Monday call, the founder was kind about it, which was worse than if he had shouted.
The ads just stopped working. Did something change?
Something had changed. It was not the algorithm, it was not iOS, it was not the auction. What had changed was that the audience had seen my one good idea enough times. I had a creative problem and I had been treating it as a media-buying problem for eighteen months, because media buying is the part I am comfortable in.
What I tried first, and why it failed
- Volume by brute force. I told myself I would ship fifteen new creatives a week. I shipped nine one week, four the next, and zero the week a client had a launch. One person cannot both run the accounts and be the creative department, and the accounts always win because they are the thing on fire.
- Generic AI writing tools. They wrote fluent ad copy about nothing. The tool did not know which of my hooks had ever worked, had never seen the account, and would happily give me a perfect paragraph about 'radiant, confident skin' - the exact sentence a hundred other brands were already running.
- Copying competitors by hand. I kept a swipe folder from the ad libraries. It was genuinely useful and completely unsustainable - scrolling, screenshotting, trying to guess from the outside which of those ads was actually spending money and which was a corpse somebody forgot to turn off.
That third failure is the one that stuck with me. I did not need more ads. I needed a faster path from evidence to idea, and every tool I had was either evidence with no ideas or ideas with no evidence.
The turning point
I found Atria at tryatria.com after a media buyer in a paid community mentioned it twice in one week, which is usually how these things reach me. I opened the homepage expecting the usual: a purple gradient, a promise to 10x my marketing, a demo request form. What I got instead was a line that made me stop scrolling, because it described the exact gap I had been failing to close.
You know what worked. Raya knows why.
Raya is their AI teammate, and the homepage's pitch for her is not shy: 'Meet Raya. Your AI teammate trained on $9B+ in real ad spend.' I want to be careful here, because that is their claim about their own training data and I have no way to audit it. But the framing was the thing. Not 'AI writes your ads', which I had already tried and hated. Instead: you already have performance data, and the useful machine is the one that reads it and tells you what the pattern is.
The rest of the page is organised the way the job actually runs, which I appreciated more than I expected: 'From insight to ads, on repeat', then 'From assets to live ads in a click', then 'Win bigger every time. Raya never loses learnings.' Research, production, launch, learning. That is the loop. My loop had been broken in the middle for two years.
Two smaller things made me willing to actually sign up rather than bookmark it and forget. The homepage says 'Start for free', 'No credit card required' and '1,000 free credits included', so I could put my hands on it before having a budget conversation with anybody. And it says 'SOC 2 type II certified', which matters when the thing is going to be connected to client ad accounts that are not mine to be casual with.
How I actually use Atria
Here is the honest version of my week now, in the order the work happens. I am describing the capabilities the product advertises and how they map onto my job - not a feature list I copied, and not a performance guarantee.
Before I write anything: research
This replaced the swipe folder. Atria advertises real-time competitor strategy extraction - it pulls top hooks, personas and landing pages from what competitors are running - and customer review mining, which turns a brand's own reviews into ad angles and hooks.
- For me: the review mining is the underrated half. The dog-bed client has thousands of reviews nobody had read since 2023, and reviews are the only place customers describe the product in their own words instead of the brand's.
- For me: competitor extraction stopped being a Saturday activity. I stopped collecting screenshots and started collecting angles, which are the thing I can actually build on.
- For me: it moved the argument on client calls from 'I think we should try a founder-story angle' to showing where the angle came from.
The part that removes the blank page: briefs and copy
Atria's pitch here is that briefs are data-driven and built from your own ad performance, and that script and copy generation is trained on ad performance data rather than on generic marketing prose. There is also auto creative tagging - it labels creatives by hook, persona, USP and format.
- For me: the tagging is quietly the most valuable feature on this list. It is the thing that turns 'the coffee ad did well' into 'founder-origin hooks aimed at the price-sceptic persona do well', which is a sentence you can build ten more ads from.
- For me: briefs are what I send to the one freelance editor I use. Before, a brief was a paragraph in a WhatsApp message and a vague vibe. Now it is a document with a hook, a persona and a reason.
- For me: I still rewrite the copy. Every time. It gets me to a draft with a point of view, and I do the last twenty percent in the brand's voice, because I know the brand and it does not.
Images, and where I stay sceptical
Atria offers AI-generated images described as powered by performance data, plus the ability to clone image ads with AI and iterate on top performers. The cloning idea is the interesting one: take the ad that is already winning and produce variants of it, rather than starting from a blank canvas every time.
I use this for statics and for iterations of things that already work. I do not use it to replace my clients' actual product photography, and I would be suspicious of any review that told you to. What it genuinely solved for me is the middle case that used to just not happen: the fifteen small variations of a winner that nobody has time to make by hand.
Getting it live
One-click bulk upload directly to Meta. That is the whole feature and it is the one I underestimated. The gap between 'we have twelve new creatives' and 'twelve new creatives are live in the right ad sets with the right naming' used to eat an afternoon, and afternoons were what I was out of. Their homepage frames that whole section as 'From assets to live ads in a click' and claims 10x faster uploads - their number, not mine, though the direction matches what I felt.
After launch: the boring stuff that saves accounts
This is the part that would have saved me in February. Atria advertises auto-scaling of winners and auto-pausing of losers, plain-English ad grading with specific fix recommendations, and proactive alerts when a creative starts declining. There is a native Slack integration, so the alerts land where I already am.
- For me: the decline alert is the whole reason I stayed past the free credits. My expensive lesson was a slow drift I argued myself out of noticing. Something that says it out loud, in Slack, on day two instead of day eleven, is worth more to me than any generator.
- For me: plain-English grading is a client-call feature as much as an optimisation feature. 'This creative is fatiguing and here is the specific fix' is a sentence a founder understands without me translating.
- For me: I keep the auto-pause rules deliberately conservative. I want software to tell me a thing is dying. I am still the one who decides when to switch it off.
What changed, honestly
I am not going to hand you a ROAS figure, because the honest answer is that four accounts in four categories over a few months is not a controlled experiment, and anyone who presents that as one is selling you something.
What I can describe is the shape of the change. I ship more creatives than I used to, by a margin that is obvious rather than marginal, and more importantly they are different from each other - different hooks, different personas, not fifteen cousins of the same ad. When something starts to fade I find out while it is fading instead of at the post-mortem. And there is now always something warmed up behind the winner, which is the single thing that would have prevented February.
The part I did not expect: the Monday calls got easier. Not because the numbers are always good - they are not - but because I can explain what I am doing and why, with the evidence attached. 'Here is the angle, here is where it came from, here is how it is grading' is a much better meeting than 'I moved some budget around and we will see'.
Is this just making everybody's ads the same?
This is the fair objection and I sat with it for a while. If a tool trained on ad performance data tells thousands of buyers what works, does everything converge into the same beige ad?
My answer after using it: the research is shared, the raw material is not. The review mining is reading my client's reviews. The briefs are built from my client's ad performance. Two buyers using the same tool on two different brands start from two different piles of evidence. And the last mile is still a person - I have never once shipped a first draft without rewriting it, because the tool knows what performs and I know what the founder sounds like when he is being himself.
The other half of the objection is more serious and I will not talk you out of it: if you have no taste and no point of view, this will help you produce mediocre ads much faster than before. It is an amplifier. It does not install judgement.
What I can and cannot verify for you
I want this section to be the reason you trust the rest of the article. Atria's homepage makes specific numerical claims - 3x creative volume without the headcount, 2-3x faster from insight to brief, 10x faster uploads, 40% higher ROAS, 70% faster launches. I have not audited a single one of them, I have no access to the methodology behind them, and you should read them as marketing until somebody shows you the working. My own experience is directional and qualitative, which is exactly why I have kept it that way instead of dressing it up in percentages.
The same goes for the social proof on the page: a 4.9 G2 rating, 20,000+ teams, logos including Ipsy, Kitsch, Blenders, Loop Earplugs and Everyday Dose, and testimonials from named people at some of those brands. Vendor-selected testimonials are vendor-selected testimonials, however real the names are.
On money: the homepage leads with 'Start for free', 'No credit card required' and '1,000 free credits included', and points to a pricing page for anything beyond that. I am deliberately not quoting you a price, because I only read the homepage and prices move. Start on the free credits, run your own account's research through it, and have the budget conversation once you know whether the output is any good in your category.
One more thing worth knowing before you assume it is a pure self-serve product: the homepage also sells a human layer - 'Not just software. Strategy too.' - described as a strategist embedded with your team, with monthly audits, briefs and scaling support. There is also a REST API. Whether either of those is in reach at your spend level is exactly the kind of thing to ask on the demo call rather than guess at.
The new Marco
The old me measured a good week by how much time I spent inside Ads Manager. It was a terrible metric. It rewarded me for fiddling, and fiddling is what you do when you have run out of ideas and cannot admit it - it feels like working and it looks like working and it moves nothing.
The new me measures a good week by how many genuinely different ideas got in front of an audience. Some of them fail. Most of them fail, honestly. But failing with eight different angles teaches me something, and failing with eight versions of the same angle taught me nothing for a year and a half.
If your account is running on one good ad
If you have one creative carrying your spend and nothing warmed up behind it, you do not have a media-buying problem waiting for a better bid strategy - you have a countdown. That was my situation and I did not recognise it until it cost a client real money. Atria's free credits are enough to run your own research and see what it pulls out of your own reviews and your own competitors, which is the only test that tells you anything real. Do that before you pay anybody, mine included.
Disclosure, plainly: this is my own experience running four small e-commerce accounts, written after living with the tool rather than after watching a demo. Everything I could not verify, I have labelled as their claim rather than my result. If you sign up through my link, AIStackVerdict may earn a small commission at no extra cost to you. It does not buy a softer review - the weekend I wasted moving sliders is in here, and so is the fact that this thing will not give you taste.
- Marco
Community discussion
Enjoyed this story? Share your thoughts below.