Multivariate Testing vs. DCO Ads

Dynamic creative optimization (DCO) is a programmatic advertising capability that enables advertisers to customize which creative is shown under various circumstances. DCO has been a widely adopted technology for over a decade now and continues to carry its weight in the marketer’s toolkit.

‍

When using a DCO platform, common use-cases included;

• Customizing creative based on location (the ad will say ‘Best Burgers in Texas’ when running in Texas)
• Customizing creative based on retargeting (the ad will show a picture of the product that a person has viewed on the site)
• Customizing creative based on live contextual data, like the weather (the ad will show the ‘Ice Cold Beverages’ ad when it’s over 100 degrees in that location)

Additionally, most DCO platforms will also auto-optimize a campaign at scale by running automated testing under the hood and attempting to learn which ads to show more frequently.  

The Pros & Cons

Although DCO will (usually) incrementally improve performance, there are several drawbacks. To start, it involves surrendering creative control to an automated system. The learnings from the automated testing that runs under the hood are not translated as actionable insights that can be leveraged by a creative team. Most importantly, DCO is a strategy used for scaling large, evergreen campaigns rather than running structured experiments that a team can learn from.

‍

One of the core differences is ultimately that DCO is for creative scaling, and multivariate testing with Marpipe is for creative testing (remember; 10-20% of a budget should go to testing, while 80-90% of a budget should go to scaling).

‍

For DCO, the input is your (finished) ads and the output is auto-optimized ad serving based on context.

‍

For Marpipe, the input is your creative assets and the output is discovering how those assets independently influence performance.

‍

Since you need to feed ads into a DCO, the main question becomes - how do you determine what are the best ads to feed it (in a world of potentially infinite creative possibilities)?

‍

To answer this question, you need fundamental creative data for your brand about which assets (pictures, words, colors, etc) your audience loves and hates the most - this is something you can’t do with DCO, and this is what controlled multivariate testing is perfect for.

‍

Both DCO and multivariate testing are effective tools for any marketer and work well in unison. Before placing inputs into DCO, it's important to build your own base of creative intelligence from testing to discover which assets should be the inputs to a scalable DCO campaign.

‍