How data driven marketing enables you to make quality decisions quickly and efficiently

Artificial intelligence and synthetic data are becoming more and more common in marketing. And they are a common part of advertising campaigns. What does synthetic data or creative testing bring to them? Barbora Gabrižová and Ondrej Dúžik, experts from GroupM, part of the WPP communications group, talk.

The interview was published in Stratégie magazine. Author: Lucia Ležovičová

Is synthetic data the reality of marketing campaigns or are we still in the realm of theories?

Barbora: Compared to other industries, such as healthcare or finance, marketing has the advantage of not being so regulated. But at the same time, it works with large amounts of data. It is therefore an ideal candidate for the rapid adoption of various technological innovations, which currently include AI and synthetic data.

Ondrej: Depending on how specific we are in defining synthetic data, we can evaluate its current use. Synthetic data simulates real data without exposing sensitive information, while trained AI models analyze existing - real or synthetic data to predict trends, simulate research results, and optimize campaigns. Synthetic data is thus only a subset within the complex AI ecosystem.

For several years now, we have seen even the largest research companies, such as Kantar and Nielsen, working to develop products that leverage synthetic data. It is currently predicted that by 2030, AI will largely only work with synthetic data instead of real data.

What are the benefits of AI and specifically synthetic data?

Ondrej: Synthetic data is useful when real data is missing or too sensitive to use. The use of synthetic data brings with it a number of benefits - speed of results, accessibility, efficiency, anonymity, and many others. At its core, it is an extreme democratization of data and the insights derived from it. AI models in general are well suited to automate, personalize, and improve the effectiveness of marketing investments.

Have you tried out these new possibilities brought by artificial intelligence in practice for specific campaigns?

Barbora: In this respect, within GroupM, we benefit from the WPP network. We don't have to rely only on third-party tools, but we have our own operating system, WPP Open, which is just based on AI. Within it, we can use selected text, image, video and voice models in a closed and secure environment. In practice, it looks like the account manager has a project management tool, and within it, he already has pre-set prompts or interactions with the AI model for the individual steps within the campaign. Gradually, we integrate other planning and evaluation tools into this project management system, which helps the AI model to take into account as much information and data as possible available for a given campaign.

But human supervision is, of course, key. Data sources are never perfect, care must be taken on what data sources a given model has been trained on, how representative the data will be for our purposes, and whether it contains biases. So in that respect, I see it as if we've each been provided with a handy assistant. With emphasis on the word “assistant”.

What other risks do you see in using these tools? How do you address the issues of ethics, transparency and credibility of their sources?

Barbora: Here too we can rely on the background and support of WPP, which has a dedicated legal team. In addition to various trainings and regular monthly calls, we have a “semaphore of tools” at our disposal. Within the framework, the different tools are evaluated from a legal, ethical and qualitative perspective. This format also helps me personally to frame an overall mindset on the topic of AI tools and ask relevant questions- what data was used to train the model? How was this data collected and what ownership does the manufacturer of the tool have over it? Who has the right to use the outputs from a given model, and how? What happens to the data that the user puts into the tool, either in terms of security or in terms of further processing?

The topic of legal implications is very dynamic, in proportion as usage grows. However, key principles include not inserting client or personal data anywhere until we have clear and satisfactory answers to the questions mentioned above. For personal data, we need to keep in mind the GDPR provision where every user has the right to ask for an account of where their data is located, a justification of why that data is being processed, and also the right to ask for it to be deleted. However, it is not at all easy to comply with this provision with an AI tool.

How can synthetic data help in optimising advertising campaigns in general?

Barbora: As it is known from marketing theory, the effectiveness of a campaign is a combination of media planning and creativity. Different sources may differ on specific numbers, but simplistically I would venture to say that half of the result is media and half is creative. Synthetic data, and AI in general, can help us improve both of these components of a campaign.

In media planning, they help us in predicting campaign performance, in optimizing it, in targeting and personalizing ads more accurately. For online campaigns in particular, machine learning algorithms have been used for years and are the basis of AI models. I will now turn to the topic of ‘attention’, which has the ambition to become a single currency for campaign optimisation, especially given the possibility of cross-media comparison. It is AI that has the potential to connect and fill in missing data for a comprehensive view across media types, including human attention modelling, in an efficient and scalable way.

What about the creative part of the campaign, what opportunities does Ai oofer in this regard?

Barbora: In addition to the actual creation of visuals, AI can help test creatives. One option is to create AI personas based on segmentation data, and then use those personas to style a focus group. For example, I personally created a persona of a young teenage girl and then asked her various questions about a planned marketing campaign. She was very persuasive, behaving exactly as you would expect a teenage girl to behave. She even chose to ignore the data sources she was supposed to answer based on because she has “a mind of her own” and knows what she wants (laughs).

But there are also more exact methods available, based on synthetic data. These include, for example, heat mapping as an alternative to eye tracking studies. We also offer such solutions to our clients within GroupM.

What results has AI ad testing produced? Did you find anything fundamentally surprising? What is the focus of AI in this area?

Ondrej: Since the beginning of the year, we have already completed several projects within the mPredict product, which is based on analyses using the international Neurons platform and supplemented with AI model outputs from the WPP Open platform. The actual findings or recommendations that the model evaluates might not be surprising to experienced creatives and neuroscientists. For example, that attention is better attracted to people and their faces, versus other objects; simple visuals are more effective for better remembering the main message; fast moving objects and their abundance fragment the viewer's attention and increase the cognitive demand to process the information we want to convey with the ad. What is surprising, however, is the speed with which we can provide such findings (hours, a few days at most) and the accuracy with which we approach real eye-tracking studies - 95%. Last but not least, of course, the biggest benefit is cost-effectiveness.

How do your clients approach this option - do they have questions, concers, mistrust?

Barbora: AI is a powerful buzzword. Many clients turn to us with questions about the use of AI in campaigns - they follow global trends and want to know what are the real possibilities of application on the Slovak market. Expectations are high, especially for speed and efficiency. At the same time, our clients contact us with requests for training and education in this area. It is a new and dynamic topic, we also see the need to set the “same language”, as we know that the definitions of AI are very diverse.

Why do you believe in Data Driven Marketing at GroupM? 

Ondrej: I will mention a well-known idea, which has been discussed in various forms by Lord Kelvin and Peter Drucker, that without quality data we cannot make quality decisions: ‘If you can't measure it, you can't improve it’ / ‘If you can't measure it, you can't manage it’. For us, we want to take advantage of the opportunities we have and make these extra quality decisions quickly and efficiently.