The future of media planning powered by artificial intelligence

There are few industries where technological developments are as visible as marketing. Compared to other sectors such as healthcare or finance, it has fewer regulations, but at the same time it deals with large amounts of data. As a result, it is still a highly creative field, which is what fuels the rapid adoption of new technologies.

The latest buzz-word is synthetic data, which is one of the many practical applications of AI. It is currently predicted that by 2030, AI will largely only work with synthetic data, instead of real data.

What is synthetic data?

Synthetic data, as defined by TechSonar from EDPS (European Data Protection Supervisor), is artificially created data generated from real records using a model that is trained to reproduce the characteristics and structure of the original data. This means that the synthetic data and the original data should give very similar results for the same statistical analysis. The degree to which the synthetic data is an accurate proxy for the original data is a measure of the usefulness of the method and model.

The generation process, also called synthesis, can be performed using a variety of techniques such as decision trees or deep learning algorithms. Advanced machine learning methods GANs (Generative Adversarial Networks) are used. GANs have been introduced recently and are commonly used in the field of image recognition. They generally consist of two neural networks that iteratively train each other.

How to use synthetic data?

The use of synthetic data brings with it several benefits - speed of results, availability, efficiency, anonymity, and many others. At its core, it is an extreme democratization of data and the insights derived from it.

Where specifically do we see room for the use of synthetic data? For us at GroupM, but also in data-driven marketing in general, we see applications in the following areas, for example:

Campaign optimisation and better ad targeting

Synthetic data allows you to simulate the performance of campaigns before they are launched. We can model different scenarios, such as how conversion rates will change for different budgets, channels or target groups. Such simulations save time and money because they allow us to optimize campaigns based on accurate predictions.

Simulating seasonal trends: AI can generate synthetic data that takes into account historical seasonal consumer behaviour, allowing you to better plan campaigns during Christmas or the summer holidays.

At Nexus Media Solutions (formerly Xaxis), we have been developing and deploying Media Optimizations (formerly Copilot), a tool that uses AI to optimize programmatic campaign buying, for over 7 years.

Consumer surveys

Traditional market surveys are often costly, time-consuming and dependent on the availability of respondents. Synthetic data allows to simulate the behaviour of different target groups based on historical data and predictive models. For example, if a company wants to analyze the behavior of young consumers aged 18-24, it can generate synthetic data that realistically represents this demographic group. Another interesting example was shared a few months ago by Mark Ritson - Synthetic data suddenly makes very real ripples. Try to guess which graph is from a real survey, and which was generated based on synthetic data.

Testing creatives with AI

Before launching an ad campaign, it's crucial to test how the creatives will resonate with your target audience. In this case, AI models help in testing different variations of ads. For example, AI can generate synthetic responses to different visual or text elements of an ad, allowing creatives with the most potential to be identified.

We don't have to go far for an example - we at GroupM, in collaboration with the Neurons platform, are already testing creatives within the mPredict product, as a complement or complete replacement for creative pre-tests. The results are visualizations called heatmaps, which we then interpret with respect to the campaign objectives.

What is the future of synthetic data in marketing?

For all the benefits that synthetic data offers, it is important to remember that it also brings challenges. The final decision will be the result of the interplay between humans and technology. Therefore, from the data user's perspective, it is crucial to check on what data sources the model has been trained on. This helps us to estimate how representative the data will be for our purposes and whether it contains biases. It also helps us to assess the risk of whether the model has been trained on illegally obtained data.

Ethics and transparency in general is a separate chapter - what limits should be set when using synthetic data to influence consumer behaviour? How to be transparent with your customers and not lose their trust? How not to lose authenticity? These are all questions that the marketing community will have to address in the coming period.

Where can you learn even more?

We recommend a look at the synthetic data from our colleague Phil (How to use synthetic data for enhanced ad addressability – Phil Tolliday, GroupM).  

You can also find a structured view in the article and video from SAS (What is synthetic data? And how can you use it to fuel AI breakthroughs?). 

Ján Hudák (Kantar Slovakia) and Rasto Kočan (Go4insight) summarized how consumer surveys are changing and how AI influences them in an interview for Strategies – Prinesie AI prieskumy bez respondentov? Syntetické dáta už dnes nahrádzajú ľudí. 

Sources: Neuronsinc / edps / Vedátor / Medium / Grandview