Dynamic creative optimisation (DCO) is one of those Meta features that sounds like it should be on by default — and sometimes it is, which is part of the problem. At Choco Media, we use dynamic creative optimisation selectively, and when we turn it off, there’s always a reason. This post covers what DCO actually does under the hood, the conditions where it earns its keep, and the signals that tell you it’s time to cut it loose.
This is written for marketing teams and founders running Meta campaigns who want more than the platform’s default recommendation. You’ll leave with a clear mental model of how Meta’s algorithm assembles and ranks creative combinations, a checklist for when DCO helps versus hurts, and the practical steps to audit whether it’s working in your account.
Whether you’re spending €500 a month or €50,000, the decision to use dynamic creative optimisation is not one-size-fits-all. Let’s get into it.
What dynamic creative optimisation actually does
DCO lets you upload multiple versions of each creative element — headlines, primary text, images, videos, call-to-action buttons — and tells Meta’s algorithm to find the best-performing combinations for different audience segments. Instead of you choosing one finished ad, you hand the system a set of components and let it assemble and test them in real time.
Meta’s ad delivery system evaluates combinations using signals like click-through rate, landing page views, and conversion events. Over time it shifts spend toward the combinations that are converting, and away from the ones that aren’t. The process is dynamic — it adjusts as audience behaviour changes, not just once at the start.
- You can upload up to 10 images or videos, 5 headlines, 5 primary text options, and 5 descriptions per dynamic creative ad.
- Meta determines which combinations each person sees based on predicted affinity, not random rotation.
- The winning combination is not global — different segments may see different winners simultaneously.
- Reporting shows asset-level performance, but Meta does not expose the full combination-level data in most reporting views.
This is meaningfully different from A/B testing, where you control exactly what runs against what. DCO is more like handing your split tests to a system with strong prior knowledge about your audience and letting it optimise continuously.
When DCO works well
DCO performs best when a few conditions line up. Understanding them helps you decide whether to reach for it or skip it.
You have genuine creative variety
DCO is not a shortcut for lazy creative work. If your five headline variants are essentially the same sentence rephrased, the algorithm has nothing interesting to test. The feature earns its value when your assets are meaningfully different — different emotional angles, different proof points, different visual treatments — and you genuinely don’t know which will land best with your audience.
Your audience is large enough
DCO needs volume to learn. In our paid media work, we typically see reliable optimisation kick in when the audience is above 500,000 and the ad set is generating at least 30–50 conversion events per week. Below those thresholds, the system doesn’t have enough signal to make confident combination choices, and you can end up with a winner picked on statistical noise.
You’re in the exploration phase
Early in a campaign — especially with a new product, offer, or audience — DCO is a useful tool for shortlisting creative directions without running a structured A/B test. You learn which angles the algorithm favours, then you can invest in polishing those directions for a more controlled phase.
- Use DCO to narrow from 10 ideas to 2–3 strong performers.
- Then build standalone ads from those winners for your scaling phase.
- Treat DCO output as research, not as a permanent media strategy.
When DCO causes problems
The honest assessment is that DCO introduces as many complications as it solves if you reach for it in the wrong context. These are the patterns we see most often.
When you need creative control for brand consistency
DCO assembles combinations dynamically, which means you can end up with pairings you didn’t approve — a headline written for one offer alongside a visual built for another, or a bold claim next to a soft lifestyle image that contradicts it. For clients with strict brand guidelines or regulated industries, this is a real risk.
If your brand team needs to sign off on every ad before it runs, DCO is incompatible with that workflow by design. You can mitigate it by keeping your asset variations tightly scoped, but you’re still trading control for convenience.
When your conversion volume is too low
This is the most common mistake. An ad set generating 10 purchases a week doesn’t have the signal to distinguish meaningfully between five creative combinations. Meta will pick a winner, but it may be picking on the basis of 2–3 conversions, which is not a reliable signal. You end up with false confidence in a combination that happened to get lucky early.
“We’ve audited accounts where DCO had confidently allocated 80% of budget to one combination based on four purchases. The ‘winner’ was an artifact of the learning algorithm, not a genuine performance gap.”
When you’re scaling a known winner
DCO’s job is exploration. Once you know what works, you don’t need the algorithm testing other combinations — you need it pushing budget into the winning creative. Keeping DCO on during a scaling phase means some portion of spend continues going to lower-performing combinations even as the algorithm nominally optimises. A clean standalone ad using the proven creative is almost always more efficient at scale.
- Scale phase: turn DCO off, use your best-performing combination as a dedicated ad.
- Exploration phase: DCO can stay on.
- Review and reset: every 4–6 weeks, check whether you’re still exploring or scaling.
How the algorithm decides what to show
Understanding the mechanics helps you predict DCO’s behaviour rather than just observe it.
Meta’s ad delivery system assigns each person an estimated action rate — a probability that they’ll take the conversion action you’ve optimised for — for each combination. It then balances that probability against the cost of showing the ad (determined by auction dynamics) to maximise the expected value per impression.
Early in the campaign
The algorithm starts with limited data and distributes impressions relatively broadly across combinations. You’ll see a mix of combinations getting exposure, and performance data will look noisy. This is expected — resist the urge to make creative changes in the first 7–10 days.
As data accumulates
The system narrows. Combinations that underperform get fewer impressions. Top performers get progressively more. The speed of this narrowing depends on your conversion volume: high-volume ad sets converge quickly, low-volume ones stay noisy longer.
What you can and can’t control
You can set rules to prioritise certain assets within DCO — Meta calls these “asset customisation” settings — but in practice the algorithm overrides preference signals when conversion data is strong enough. If you want to guarantee a specific headline runs, the only reliable method is a standalone ad.
- You can see asset-level performance in Ads Manager under the “View chart” option per ad.
- Combination-level data is not available in the standard interface — you can pull it via the Ads Reporting tool.
- Assets with zero impressions after 5+ days are likely being suppressed by the algorithm; consider replacing them.
DCO and our paid media audit process
When we review an account, DCO usage is one of the first things we check. The flag isn’t whether it’s being used — it’s whether it’s being used in the right context. We look for four patterns:
First, DCO running on low-volume ad sets where the conversion data can’t support reliable learning. Second, DCO still active on ad sets that are clearly in a scaling phase with a dominant creative winner. Third, DCO with creative assets that are too similar to generate useful differentiation. Fourth, DCO with no reporting review — campaigns where nobody has checked asset-level performance in 30 or more days.
If you want a full walkthrough of how we structure account reviews, the paid media service page covers what we look at and how we prioritise fixes.
The 15-minute DCO check
If you want to audit your own account quickly:
- Open Ads Manager. Filter for active ad sets using dynamic creative.
- For each, check weekly conversion volume. If below 30 events/week, flag it.
- Click into the ad, select “View chart” to see asset-level performance. Is one combination getting 80%+ of impressions? If yes, assess whether you’re in exploration or scaling mode.
- Check the assets themselves — are the variants meaningfully different, or are they minor rephrases of the same message?
- If scaling: duplicate the ad set, use the top combination as a standalone ad, pause DCO version, compare performance over 7 days.
DCO versus Advantage+ Creative
Meta has been pushing Advantage+ Creative as the evolution of DCO — it goes further by letting the algorithm apply its own creative enhancements (background generation, image cropping, text overlays) rather than just mixing your uploaded assets.
Our assessment: Advantage+ Creative is useful for certain DTC contexts where creative volume is the bottleneck and brand constraints are loose. For most of the B2B and service-sector clients we work with, the loss of control outweighs the convenience. We typically enable individual Advantage+ enhancements selectively — music for video, for instance — rather than turning on the full suite.
For a closer look at how we structure creative testing across paid channels, the AI content creation service covers how we build the creative inputs that feed these systems.
- Advantage+ Creative = algorithm modifies your assets in addition to mixing them.
- Standard DCO = algorithm mixes your assets as-is.
- Manual ads = you specify exactly what runs.
- The right tier depends on your creative volume, brand constraints, and conversion data.
Practical recommendations
Based on what we see across client accounts, here are the defaults we’d recommend:
Use DCO when you’re entering a new audience or testing a new angle and genuinely have 5+ meaningfully different creative variants to test. Set a review calendar — we do it every 4 weeks — to assess whether you’re still learning or now scaling. Turn DCO off and move to standalone ads once one combination has clearly pulled ahead and conversion volume is sufficient to trust the signal.
Don’t use DCO as a set-and-forget system. The algorithm converges, but it doesn’t stop spending on suboptimal combinations entirely — and your cost per result will drift upward over time if you’re not actively managing the creative cycle.
For B2B or service businesses with lower conversion volume, structured creative testing — running two standalone ads simultaneously, measuring over 2–3 weeks — often gives more trustworthy signal than DCO. It’s slower, but the data is cleaner.
If you’re unsure whether DCO is helping or hurting your current campaigns, that’s often the answer: active management and clear criteria for turning it on and off almost always beats passive reliance on the platform’s defaults. If you’d like a second set of eyes on your account, reach out and we can take a look.