"Dynamic creative optimization" now describes four different products. One shuffles your headlines. One builds an ad per product in your catalogue. One makes the creative in the first place. One tells you which element worked.
Buy the wrong one and you get 300 combinations of the same idea.
This guide sorts the categories, gives you the questions to ask on a demo, and shows what "at scale" actually looks like in live accounts.
The four things sold as DCO
| Category | What it does | Examples | Best for |
|---|---|---|---|
| In-platform combination | The ad network mixes your images, headlines and CTAs per user | Meta Advantage+ creative and dynamic creative, Google Performance Max | Everyone. It's free and already in your account |
| Feed-based DCO | Turns a product feed into one ad per product, from templates | Hunch, Smartly, Creatopy | Ecommerce with big catalogues |
| Creative automation / versioning | Mass-produces sizes, languages and market versions of one concept | Celtra, Storyteq, Bannerwise | Brands rolling one campaign across markets |
| Generation platforms | Make the creative itself: scripts, statics, UGC-style video, then publish | Marxx | Teams whose bottleneck is ideas and production, not resizing |
A fifth category, creative analytics (Segwise, Motion), doesn't make ads at all. It tags them and tells you which hook or format worked. Useful, and often bought alongside.
Start with what's free
Meta's own dynamic creative lets you upload up to 30 components per ad: 10 images or videos, 5 primary texts, 5 descriptions, 5 headlines, 5 CTA buttons, plus your URL. Meta assembles and serves the combinations, and Advantage+ adds automatic enhancements like image expansion or music.
For most advertisers that is the DCO layer. You don't need to buy it.
What it can't do:
- Make a new creative idea. It only recombines what you upload.
- Tell you why something worked in a way you can brief from. Asset-level reporting is thin.
- Build the 20 versions of one concept that a multi-city or multi-language brand needs.
What recombination looks like
Supply6 sells an all-in-one nutrition shake. Here is the base ad: the product, the promise, and nothing else competing for attention. It is a clean starting point, and importantly it is the only one of the three that required somebody to decide what the ad was about.
Now the same product with the elements rearranged. New layout, same argument, same pack shot. This is precisely the kind of variation a combination engine produces on its own, and Meta will happily serve it to whichever audience responds to it. Nobody had to think about a new customer, a new moment or a new reason to buy — the inputs were already in the account.
And a third time, with a discount sticker and a promo code bolted on. Three live ads, one idea, three arrangements. If you were counting variants you would report three; if you were counting arguments you would report one, and the second number is the one that decides whether the test can teach you anything.
That is what the free layer is for, and it does it well. What it cannot do is decide that the next ad should be about something else entirely.
That's the honest gap the paid categories fill.
What "dynamic at scale" really looks like
Most DCO pages don't show you the thing you actually need to see: what a winning version strategy looks like in a live account. There are two shapes, and they call for different tools.
Shape one: one offer, many versions
Foodstories runs a free-delivery offer as a separate ad per city rather than one national ad with a generic promise. This is the Bangalore cut, and it has been live 93 days — long enough to say the structure works rather than that it got lucky.
And here is Hyderabad. Same layout, same offer, same product photography; the city name is the only variable that moved. Multiply that by every city you deliver in and you have twelve or twenty live ads that are, in every sense that matters to a designer, the same file saved repeatedly.
That's a versioning job, and it's exactly what feed-based and creative-automation tools are built for: one master, many locales, a dozen near-identical copies live at once. Doing it by hand is where designers lose their week.
Shape two: one product, different ideas
Licious ran a flat 20% off for first-time buyers: a clean acquisition offer with free delivery attached, aimed at anyone who has not ordered yet. There is nothing clever about it and nothing needs to be — it is the ad that catches people already looking for a reason to try the category.
In the same window they ran this: "Retire hurt. From the kitchen." Tandoori chicken, ready in two minutes, aimed at someone whose match is about to start and who does not want to spend the first over at the stove. The product is identical to the one in the previous ad. Everything else — the buyer, the moment, the reason to act tonight rather than eventually — is different.
No combination engine would have produced the second one. Different audience, different tension, different landing page, and a joke that only works during a cricket match. That's a production job, not an optimisation job.
Keep both shapes in mind on every demo. Ask the vendor which of the two they do — most do one and imply the other.
How to choose
| Your bottleneck | The category to buy |
|---|---|
| "We have 4,000 SKUs and no ads for most of them" | Feed-based DCO |
| "One campaign, 12 markets, 40 sizes" | Creative automation / versioning |
| "We ship two new ideas a month and need ten" | Generation platform |
| "We don't know which hook works" | Creative analytics |
| "We just want Meta to test our headlines" | Meta's own dynamic creative, free |
Most teams buy one, and bolt the free layer underneath it.
Ten questions for the demo
- Does it create new creative, or only recombine what I give it?
- Can it make video, or only statics and banners?
- Does it publish straight into Meta, or export files I upload myself?
- Does it publish paused, so nothing spends before a human looks?
- How many placements does it auto-cut (9:16, 1:1, 4:5, carousel)?
- Can it keep brand rules (fonts, colours, logo, legal lines) across every version?
- Does it show which hook, offer or format won, not just which ad ID?
- Can it version by language and city without a designer?
- What's the approval flow before anything goes live?
- What does it cost at my volume, and does the price move with ad spend?
Question 1 is the one that sorts the market. Everything else is implementation detail.
Rough price shape in 2026
| Category | Typical entry point |
|---|---|
| In-platform (Meta, Google) | Free |
| Design-led automation | From roughly $45–$300 a month |
| Meta-focused optimisation tools | From roughly $99 a month |
| Feed-based DCO | Around $2,700 a month at the upper end |
| Enterprise creative platforms | Custom pricing, annual contracts |
Published prices move, so treat these as a shape, not a quote. Ask what happens when your spend doubles.
Four traps
Combination counting. "4,000 variants" sounds impressive until you notice they're all the same idea. Meta needs different ideas to have anything to optimise between.
Resizing sold as creative. Cutting one video into nine aspect ratios is useful plumbing, not a creative strategy.
Feed DCO for a small catalogue. If you sell one product, a service or a subscription, a product feed has nothing to work with — there are no SKUs to turn into rows. Nerolac is a clean example: it sells painting services, so there is no catalogue to multiply. The first ad sells finish and reliability — expert painters, perfect colours, the whole job handled — which is the pitch for someone who has been putting the decision off because of the hassle.
The second sells a season and a fear of damp. Get monsoon-ready with expert waterproofing: the same company, the same crew, talking to a homeowner who has just watched a stain spread across a wall and now has a deadline the weather set for them.
Two ads a feed could never have generated. Both ran three to four weeks. That variation came from a brief, not a spreadsheet — your variation lives in the job, the occasion and the offer.
No path back to the brief. If the tool can't tell you which hook won in words you can write a brief from, you'll be guessing again next month.
Where Marxx fits
Marxx sits in the generation category, with the research and reporting either side of it.
- Research first. It indexes over 100 million ads across Meta and Google, decodes each one into 16 attributes, and shows what competitors keep running. That's how you pick the idea before you multiply it.
- The generation canvas. One brief becomes scripts, statics, UGC-style video with AI avatars, voice cloning and lip sync, then every placement cut: 9:16, 1:1, 4:5 and carousel.
- Straight into Meta. Campaign, ad set and ad are built from the canvas and published paused, so nothing spends before you look.
- Loop closed. Fatigue scoring and hook comparison after launch decide the next brief.
In practice that means the Licious example above is one afternoon: same product, four audiences, each with its own hook, each in every placement, without a designer in the middle.
Trial the canvas: turn one brief into a finished ad
Bring one brief, a product, a customer and a moment. We'll take it from research to a finished ad on the production canvas and push it to your Meta account, paused.
Book a demo and connect one ad account on the call.
FAQs
What is dynamic creative optimization?
Serving different combinations or versions of an ad to different people and letting performance decide which wins. It ranges from a network mixing your headlines to a platform generating a distinct ad per product, city or audience.
Is Meta Advantage+ creative the same as DCO?
It's Meta's version of it. Advantage+ applies automatic enhancements, and dynamic creative combines the assets you upload (up to 30 components per ad). It optimises what you give it; it doesn't create new ideas.
Do I need a DCO platform if I use Meta's dynamic creative?
Only if you have a bottleneck it can't solve: a big catalogue, many markets or languages, or not enough new creative each month. Otherwise the free layer is enough.
What's the difference between DCO and creative automation?
DCO decides what to show each person. Creative automation produces the versions in the first place: sizes, languages and market cuts. Many tools do both, which is why the labels blur.
How many creative variations should I actually run?
Fewer combinations, more distinct ideas. A handful of different hooks and formats teaches you more than hundreds of recombinations of one concept.
Does DCO work for services and experiences, not just ecommerce?
Yes, but not the feed-based kind. With one product or a service, your variation comes from occasion, audience and offer — an acquisition offer and a match-day craving sell the same product to two different moments, as in the examples above.