The perils of presumptive personalisation

30 Jun 2017

For marketers keen to realise their goals of increased ROI and brand perception, now is the time to implement personalisation and data driven creative. There are a variety of data points available to brands and agencies to better segment audiences and ensure they receive ads that are tailored to their interests, and contain specific products, messaging or imagery that will appeal to them.

Dynamic creative optimisation (DCO) offers marketers the optimum means of personalised targeting. But the technology is still new and many marketers have found it difficult to adjust to this new paradigm and make the best use of DCO technology and capabilities. This is especially true for marketers with limited ecommerce or digital presences, such as fast-moving consumer goods (FMCG), finance and pharma.

In many cases, advertisers become overwhelmed with identifying a target audience, searching users from a media perspective, and altering messages to suit each segment. But perhaps the biggest issue is that many brands and agencies still apply traditional marketing and assumptions to DCO technology. Ultimately, this thinking limits DCO’s efficacy and marketers will have to move beyond it if they truly want the best possible campaign results. Fortunately, with a little experimentation, this is easy to do.

There is a prevailing belief that personalisation is just serving different messages to different audiences, and the optimisation element is simply serving varied iterations of creative (the colours, backgrounds, or layouts) to get incremental performance based on small aesthetic differences. When marketers first adopt DCO with very little first-party data, they often bring along rigid ideas of audience along the lines of gender, age and income, as well as vague personas, like “fashionistas” or “sportsmen”. As a result, marketers face the challenge of finding enough reach to buy these very specific segments, and feel frustrated having to rely on third-party data with varying levels of transparency and accuracy.

While these target audiences and personas are still useful and should certainly be included in a DCO strategy, the beauty of the technology is that marketers can learn so much more about what resonates with existing target audiences, as well as discovering high-performing audiences they might never have considered. Often a marketer’s initial idea of who’s worth reaching and what message resonates within each audience segment is wildly different from the DCO campaign’s outcome. Therefore, if marketers base their first campaign around presumptive personalisation, they may come out thinking that DCO doesn’t perform, rather than learning about how their product resonates with a broader audience. This is important, because that new audience can be leveraged to drive more acquisition and conversion.

When starting out with DCO, marketers should take the following three steps, especially if they are running branding or acquisition campaigns or have limited data.


  1. Develop a variety of creative concepts, focusing on varied messaging and imagery rather than minute creative changes in color or layout. Think of different product USPs, use different voices and tone, and relate your product to different subcultures and interests. This doesn’t have to be a Herculean effort; start with the minimum variations that provide the greatest difference in meaning among them.

  2. Instead of sending specific messages to a limited list of target personas, choose a variety of audiences across the funnel that may or may not already be in consideration, perhaps based on geography, context, or any other available audience data. Then, serve all of the creative variations built in step one to all of the audiences. Run these creatives using auto-optimisation towards the chosen KPI for enough time and impressions to achieve a meaningful sample size. This will likely be a small portion of the overall budget to start, but ideally marketers should look to at least a few million impressions spread over more than a month.

  1. Finally, take a look at the data and remove audiences that don’t perform at all from your plan, even if they were among the original personas. Then, double down on the audience segments that work and explore these in more granularity.


DCO rarely works as a “set it and forget it” tool, even for the most sophisticated marketers. When implementing personalisation campaigns with limited first-party data resources, it’s well worth the initial “wasted” impressions for marketers to explore outcomes beyond their presumptions. The beauty of DCO – and all data-driven targeting – is that it often reveals valuable audiences that the marketer has never even considered. It’s these unexpected audiences and insights that truly unlock DCO’s performance potential.


Source: Kelsey Meuse, Product Marketing Manager, Sizmek.