What Is Dynamic Creative Optimization (DCO)? A Guide to Making Every Variant Measurable

Dynamic creative optimization (DCO) assembles personalized ads in real time. See how DCO works, how to pick tools, and how to keep every variant measurable.

Rob Allanach11 min readPillar
Loose jigsaw pieces scattered beside pieces interlocked into a seamless grid

Dynamic creative optimization (DCO) is a form of programmatic advertising that assembles and personalizes an ad in real time, swapping images, copy, offers or layout for each viewer based on data signals such as audience, context and past performance.

A base template, a feed of modular assets and decision logic produce many variants on the fly. The leading DCO guides cover that part well.

What they leave out is what happens after the variants ship. DCO can create hundreds or thousands of versions of one ad, and unless each carries a persistent identifier and consistent metadata, you cannot reliably tell which version drove which outcome, or reuse it, across your ad platforms, your asset library and your analytics. The optimization is only as measurable as the data underneath it.

What is dynamic creative optimization (DCO)?

Dynamic creative optimization is a programmatic advertising technology that customizes an ad’s creative elements in real time for each viewer, using audience, context and performance data.

The creative elements are the pieces of an ad that can be varied: background, image, headline, body copy, offer and call to action. Google’s Studio Help page “About dynamic creatives” describes the mechanics: the creative file acts as a frame, any element in it can be swapped, including call-to-action text, exit URLs and images, and a dynamic feed holds all the content variations the creative can display, so that relevant content is served on every impression.

Optimization usually relies on multivariate testing, which compares many combinations of elements at once rather than two whole ads.

The objective can be as early in the funnel as engagement and clicks, or as late as purchases and lifetime value.

Dynamic creative and dynamic creative optimization are not the same thing. Dynamic creative swaps ad elements from a set of options. Dynamic creative optimization adds a decisioning layer that uses data, and in more advanced setups machine learning, to choose which combination to serve each viewer.

A product that rotates headlines is doing the first. A product that learns which headline works for which audience, and changes delivery accordingly, is doing the second. DCO here means the creative side of programmatic advertising: how variants are made, chosen and measured.

Name every variant onceClaravine Content applies governed tags and a persistent Content ID when creative is made.Explore Claravine Content

How does DCO work?

DCO works by combining three inputs, a base creative template, a feed of modular assets and decision logic, so the ad-serving system assembles a variant for each impression.

The template fixes the parts that never change, such as the frame, logo position and legal line. The asset feed supplies the parts that do, such as images and offers. The decision logic, rules or a model, picks one combination per impression.

Here is an illustration, not a case study. A retailer builds one template and a feed of forty products, six headlines and three offers. The serving system can now produce 720 combinations of those parts without a designer building any of them. A viewer who browsed running shoes that afternoon might see shoes with a discount.

A related pattern exists on the large ad platforms. Google’s help page on how to create a responsive display ad says advertisers supply images, headlines, logos and descriptions, Google generates ads automatically from them, and the ads can show in thousands of layouts.

The levels of DCO: from basic personalization to real-time optimization

DCO runs from simple rule-based swaps to machine-learning systems that continuously shift delivery toward the best-performing combinations.

The levels are best read as a ladder.

  • Rule-based personalization. A person writes the rules: this product in this region, this offer to returning visitors. Nothing is learned. The benefit is control; the limit is the person’s guesses.
  • Feed-driven dynamic creative. Product, price and availability come from a live feed, so the ad stays accurate without manual edits.
  • Test-and-learn optimization. The system serves many combinations, reads results and favors the winners within a campaign, through the multivariate testing described above.
  • Real-time, model-driven optimization. A model weighs signals such as time of day and recent behavior for each impression and keeps adjusting as results arrive.

Each step up multiplies the variants in market, and the identity of each one matters more as the count grows, because the learning only helps if you can read it back.

Why marketers use DCO: the benefits

The core payoff of DCO is relevance at scale: one campaign can serve thousands of tailored variants without a designer building each one.

  • Relevance. An ad that reflects the viewer’s context is meant to be more relevant than a single message sent to everyone. Whether it performs better in your account is something to measure, not assume.
  • Production efficiency. Modular assets are made once and recombined. A change to an offer is a feed update, not a re-render of every version.
  • Faster learning. Multivariate testing exposes which parts of an ad do the work, instead of only which whole ad won.
  • Adaptation to context. A fixed split test reports one average across season and geography. DCO can account for those factors.

There is a cost on the other side of the ledger. Every extra variant is another thing to name, store and report on, and a program that generates variants faster than it can describe them ends up with data it cannot interpret.

Why creative measurement breaks at DCO scale

At DCO volume, the same creative ends up labeled differently in the ad platform, the asset library and the analytics tool, so a variant cannot be tied back to its results.

The ad-serving system can name variants by its own logic, such as an internal combination ID. The asset library names files the way the creative team named them. The analytics tool groups by whatever arrived in the tracking parameters. Three systems, three vocabularies, each internally consistent.

The failure shows up in the report. A table lists “variant_4471_b” next to a spend figure, and nobody can say which headline and image that was, or whether the library asset is the one that ran. See ad trafficking for how the handoff between creative and delivery produces exactly this kind of mismatch.

In Claravine’s own work, taxonomy fragmentation across channels and systems is a recurring obstacle to unified measurement. The pattern is not specific to DCO, but DCO makes it worse because it multiplies the number of things that need a consistent label.

A platform dashboard reports on what that platform served. Whether the same asset also ran elsewhere under another name is a cross-system question, and it needs a shared key that exists before the ad runs.

What DCO depends on: a governed creative-data layer

The governed creative-data layer is a persistent creative identifier plus standardized metadata, applied when the asset is created, that makes DCO variants findable, reusable and measurable.

A persistent identifier is a code assigned to an asset once and kept for its life, whatever it is renamed to. Standardized metadata is a fixed set of fields, filled from agreed values, that describe the asset: concept, format, message, offer, market, campaign. Together they let a system recognize that two records, in two tools, are the same thing.

The industry has started to formalize the identifier half of this. The IAB Tech Lab’s standards listing includes the Ad Creative ID Framework (ACIF), which it describes as defining a framework to enable a persistent identifier for ad creative at every phase in the video advertising supply chain. It is video-focused, and it shows that identifying creative persistently is a recognized problem.

The engine builds each variant and the ad platform reports on it under its own ID. With no shared identifier there’s nothing to join the two on. You know a variant ran, not which one worked.

Kaden Carroll · Lead Solutions Architect, Claravine

The layer does not generate or score variants, and it does not prove that one creative caused a result. It makes the question answerable: when the DCO engine reports on a variant, the report can name an asset that exists in your library, described in your own vocabulary.

How to connect DCO variants to campaign performance

You connect DCO output to performance by giving each asset a persistent identifier and a standard metadata set at creation, then carrying both through serving tools into analytics.

There are four steps.

  1. Define the fields before the first variant exists. Decide which attributes you will ever want to cut results by: concept, format, message, offer, audience, market. Keep the list short. Naming conventions covers how to write values that hold up.
  2. Assign the identifier at production. Recording it then is easy, and reconstructing it later is harder.
  3. Pass the identifier into the DCO and ad-serving tools. Whether a given tool carries a custom ID through to its reporting is a question for that tool. Ask it before you buy, and test it with a small batch.
  4. Join on the identifier in analytics. Performance then attaches to the asset and its attributes, not just the campaign.

Claravine Content covers the first three. It assigns a persistent Content ID to each asset at production, attaches attributes from your own taxonomy, and carries the identity out of the asset library through the workflow tool and into the ad platforms, per its product page. It connects to the library you already run rather than replacing it. The fourth step happens in the warehouse or analytics tool you already report from.

One customer, a Fortune 100 retail company’s media science team, described the foundation this way.

“Our big task on the media science team and responsibility in our early years are over data standards and governance. Claravine helped lay that foundation…helping to increase paid media tracking by 65%.” — Fortune 100 Retail Company, media science team

That figure is about paid media tracking at campaign level. It is not evidence that any individual creative variant out-performed another, and it should not be read as such. For the measurement design that sits above this, see marketing measurement.

How to evaluate DCO tools, platforms, and vendors

Evaluate a DCO platform on whether its output stays governable, as well as on variant volume, channel reach and optimization method.

Those are real criteria and you should apply them. Add a second set, which decides whether you can learn anything from the first.

  • Naming and ID control. Can you set your own variant or asset identifier, or does the tool only generate its own? Can that identifier be exported and appear in reports?
  • Taxonomy at ingest. Does the tool accept metadata from your standard when assets arrive?
  • Library integration. Does it read from your asset library, and does the asset’s identity survive the trip?
  • Reporting at asset level. Can you export results by the elements used, such as headline and image, and not only by the combined variant?
  • Rights and expiry. Can an asset be pulled from rotation when its usage rights end?
  • Optimization transparency. Can you see why the system favored a combination?

Whatever the vendor category, whether ad-serving platform, creative-automation tool or an ad network’s own dynamic format, each can be asked the six questions above. Buying on volume alone risks finding the identity problem when the first cross-channel report cannot be built. Settle the identifier and the fields first, then judge each tool by whether it respects them.

Getting DCO adopted across brand and agency teams

DCO programs scale when governance travels with the creative, so agencies and internal teams keep working in their own tools while shared standards apply themselves.

A naming document does not do this. The workable approach is to make the compliant path the easy one: the values on offer at the point of work are the permitted values, and a record with an empty required field is stopped.

This is where DCO programs can stall. The creative team, the media agency and the ad-tech vendor each have a tool, and each tool may have a different name for the same variant. Asking all three to adopt a new process is hard. Asking none of them to change tools, and connecting the identifier underneath, is easier.

Colgate-Palmolive’s Global Head of Ad Operations, Measurement, and Data Integrity described the effect on agency teams.

“With these integrations in place, our agency teams can confidently set up campaigns, knowing that every element aligns seamlessly with Colgate’s taxonomy requirements.” — Eric Kirtcheff, Global Head of Ad Operations, Measurement, and Data Integrity, Colgate-Palmolive

That quote concerns campaign setup, not DCO specifically; the relevance is that alignment happens in the tools people already use.

Give every variant an ID you controlSee how Claravine Content carries one persistent Content ID from production to reporting.Explore Claravine Content

Frequently asked questions

What is dynamic creative optimization (DCO)?

DCO is a form of programmatic advertising that builds an ad for each viewer from a base template, a feed of modular assets and decision logic. The aim is the most relevant combination of image, copy, offer and layout for each impression.

How does dynamic creative optimization work?

The serving system takes the template and the asset feed, picks a combination for each impression using rules or a model, and records the result. Over time, optimization shifts delivery toward the combinations that perform best against the objective you set.

What is an example of dynamic creative optimization?

A travel brand might show the same base ad with a different destination, price and image depending on the viewer’s location and recent searches. This is an illustration, not a client result.

What should I look for in a DCO platform?

Look beyond variant volume and channel reach. Check whether you can set your own asset identifiers, ingest metadata from your taxonomy, connect to your asset library and export results by element, because those decide whether you can learn from the output.

How do you measure DCO performance across so many variants?

Give each asset a persistent identifier and standard metadata when it is made, carry the identifier through the serving tools, and join on it in analytics. Without a shared key, each system reports on its own names and the results cannot be reconciled.

How do you get agencies to follow DCO governance without adding friction?

Put the standard inside the tools they already use, so permitted values are what they pick from and the identifier is applied automatically. A process outside the workflow is easy to skip.

Why can’t we match ad results across channels when each platform names creative differently?

Each platform reports on its own labels, so one creative appears under several names and no shared key joins the rows. The fix is a single identifier and a standard set of tags assigned before the ad runs, then carried into every system. Claravine Content applies your taxonomy when creative is made and gives each asset a persistent Content ID that travels with its metadata into ad platforms and analytics.

How do we keep variant names clean when DCO assets go to the ad platforms?

Variants multiply quickly, and each one needs a name that matches the campaign taxonomy when it is trafficked. A typo or a missing field sends a variant out unmatched or blocks the upload. Generate names from approved values and a variant key instead of typing them. Claravine Content applies the taxonomy when each variant is created and attaches a persistent Content ID, so the metadata is already correct at trafficking.

Sources

Outbound citations, named and dated:

  • Google Studio Help, “About dynamic creatives” — the creative file as a frame, swappable elements, the dynamic feed, relevant content on every impression.
  • Google Ads Help, “Create a responsive display ad” — advertisers supply assets, Google generates ads automatically, ads can show in thousands of layouts.
  • IAB Tech Lab, “IAB Tech Lab standards listing” — Ad Creative ID Framework (ACIF) as a persistent identifier for ad creative across the video advertising supply chain.

Related Posts

Free guideThe 2026 Healthcare Marketing Data Compliance ChecklistA 20-point audit for healthcare marketers to self-assess compliance risk in their marketing data — consent, campaign tracking, vendors and internal governance under HIPAA, GDPR and state privacy laws — plus four warning signs you're likely non-compliant.Get the guide