Magic Numbers: Asking The Right Questions with Eric Walzthöny
Eric Walzthöny Kreutzberg, co-founder and chief technology officer, blends a wealth of scientific academia — from mechanical engineering, to biotechnology and AI — with a career based on research and data science. With his fervour and steadfast energy to bring media and creativity closer together, Walzthöny has led a team of developers and engineers to bring ad tech platform SmartAssets to market.
Over the past 18 months, his team has designed and delivered complex software architecture that analyses the creative components of assets at granular level, scoring them in-line with best practices and providing recommendations to drive creative effectiveness. As the AI space continues to develop, Walzthöny and the SmartAssets team are also planning new functions that will allow instant in-platform optimisation.
LBB> What’s the number one question that clients are coming to you with when it comes to how they can better use data to enhance the creativity of their content and experiences?
Eric> The question that we get asked most frequently is, “how can we leverage the data that we already have?”
And this is a great starting point. Instead of seeking insights from current data — which isn’t always easy and can be a little counterproductive for campaigns that are already active — we’re hearing a lot more questions around historical data.
Creative leaders are seeking to understand how data can inform and stimulate creativity, but they understandably don't want to be dictated to.
LBB> How can you make sure that data is elevating creative rather than forming a wind tunnel effect and knocking all the interesting or unique edges off that make something distinctive?
Eric> We need to use data as a guide, rather than a rule.
Essentially, what this means is using data-led insights to spark creative ideas, instead of stubbing them out. To this effect, SmartAssets offers different levels of testing to validate creative brainstorming.
This could take the form of standard A/B tests which essentially look at creative variations and gauge whether a particular idea resonates with its intended audience, or whether it’s best to kill it off early and save the media spend preflight.
If a brand is looking to go a step beyond that, we’d go more in depth with incrementality testing and underscore what’s really working in an entire repository of ads. This consists of hooking up to a client’s DAM and, if possible, ad platforms. We then analyse granular creative components and correlate them with performance data on chosen ad platforms such as Meta, Snapchat, TikTok, etc.
It’s important to remember that this is an iterative process, which means incorporating data and feedback throughout the entire creative process, and allowing for adjustments that maintain a distinctive edge. Creatives should feel supported by all this data which will enable them to tap into valuable insights, channel creativity and ultimately bolster effectiveness.
LBB> Can you share with us any examples of projects you’ve worked on where the data really helped boost the creative output in a really exciting way?
Eric> For a recent client project, we analysed the performance data for a series of static ads and videos.
We extracted in-house metrics, including information on engagement, potential, attention, cognitive load, memorability, and various other data points all the way down to more abstract rules such as human priming (the process of triggering memories).
We put this into our machine-learning pipeline, extracted the attributes, and predicted the target KPIs for campaigns with a specific objective to increase online sales in several markets.
Using proprietary platform data and machine learning, we were able to find a sweet spot for online-sales-boosting elements such as text, font, logo sizes, emotion, product quantities, etc. What we found is that when implementing these recommendations, each campaign’s results would be boosted by 28%.
LBB> More brands are working to create their own first party data practice - how can a brand figure out whether that’s something that is relevant or important for their business?
Brands can determine the relevance of first party data by assessing their direct customer interactions and data needs. This could include evaluating the data sources that they have, such as existing customer touch points, and looking at other data collection opportunities.
So starting off, you could look at the investment level required to get the data, and compare this figure against the potential insights and ROI. And once you’ve identified the value of the data, you need to make sure that there's strategic alignment. Essentially, this means ensuring that the data practices that you've worked on align with the overall business and marketing strategies.
The data collected should be important, efficient, and affordable.
To illustrate this, the most important data that we're collecting is the interaction of the end user with ads. This includes digital metrics such as clicks, views, through rates, time spent interacting, etc. But we also go a leap further than that and take it to a cognitive level, measuring interactions through proxies of published consumer research.
These are like clinical trials which use precise apparatus which are able to denote real-life measurements such as reaction type, attention focus, EEGs (electroencephalograms), etc. We have a lot of good proxies in place from machine learning models, which essentially just replicate the way our brain works.
LBB> We talk about data driving creativity, but what are your thoughts about approaching the use of data in a creative way?
Eric> Data is absolutely everywhere, and I’m constantly on the lookout for unexpected patterns and insights that can drive innovation. So, in the context of using data creatively, we can break this down into three different parts:
The first part is carrying out exploratory data analysis, which just means a deep dive into the data sets without having any bias or preconceived notions.
The second part is trying to find the story that the data is telling us, and then working out how to convey this story to somebody that is otherwise unable to interpret the data.
There's very much an emotional appeal to this aspect — presenting the data in a compelling way, and ensuring that it resonates with the end users.
Lastly, we focus on collaborative interpretation by actively engaging with our marketing and product teams. While our technical expertise allows us to extract valuable insights, we recognise that integrating perspectives from marketing and product is crucial to fully understanding and leveraging the data. By working together across functions, we can gather feedback, view the data through different lenses, and ultimately foster more diverse and creative outputs that align with our broader goals.
LBB> "Lies, damned lies, and statistics" - how can brands and creative make sure that they’re really seeing what they think they’re seeing (or want to see) in the data, or that they’re not misusing data?
Eric> There's always this classical notion of 'bad data in, bad results out'. So, first and foremost, we need to make sure that we have robust strategies to collect strong data. That way we can assume that the insights are on point from the get-go,
But on top of that, we also need to avoid potential stumbling blocks by setting up a QA process for data processing. This means using robust analytics tools, in combination with tested methodologies which are reliable and valid when confronted with any sort of doubt.
There's always a caveat of presenting the data in a different way, but that is ultimately the choice of the person who is presenting the data. You can have a great process for collecting data. You can have a great team of scientists and statisticians to value your data. But at the end of the day, you can still manipulate the data to say what you wanted to say, right?
This is why we should always be focusing on cross-verifying data and validating findings with multiple data sources. We also need to educate teams that are going to interpret this data for specific techniques, to avoid biases or misinterpretation.
LBB> What are your thoughts about trust in data - to what extent is uncertainty and a lack of trust in data (or data sources) an issue and what are your thoughts on that?
Eric> It’s a bit like a relationship, where you eventually build trust over time.
When working with data you have to validate the source, be transparent, and have rigorous validation steps — looking at where it comes from, documenting it, sharing its origin, and processing it throughout various stages. This way, if somebody notices something is odd when cross-referencing the data, they can raise questions, adjust the documentation, and we adjust how it’s processed.
The lifecycle of this data is incredibly important, and what a lot of marketers don’t know is that in order for data to be clean, the noisy elements are removed. Yet it’s these noisier elements that contain interesting patterns, or golden nuggets, which are often lost.
So, keeping track of unspoiled data for new breakthroughs in technology — like what SmartAssets is doing with automated granular creative tagging — is crucial.
We have a responsibility to put robust guidelines in place for good sample data collection, while also ensuring that our data represents an even and comprehensive demographic.
LBB> With so many different regulatory systems in different markets regarding data and privacy around the world - as well as different cultural views about privacy - what’s the key to creating a joined up data strategy at a global level that’s also adaptable to local nuances?
Eric> Flexibility is key.
To navigate different regulatory systems and cultural views, we must meet local data regulations — whether it’s GDPR, CCPA, or other compliance acts — while avoiding a rigid, one-size-fits-all strategy. For example, both Spain and Germany operate under GDPR, but Germany's stricter cultural stance on data requires a more tailored approach.
A successful global strategy should establish a flexible framework adaptable to local nuances, supported by insights from local experts. Continuous monitoring, rapid adaptation, and technological agility are essential to ensure compliance and responsiveness across diverse markets.
LBB> What does a responsible data practice look like?
Eric> There are four steps to responsible data practices: transparency, user consent, data minimisation, and security measures.
Transparency - this is the clear communication about data collection and usage, which is also required by GDPR.
User consent - the user must understand what they are consenting and agree to their data being collected.
Data minimisation - only collecting the data that is pertinent to a specific purpose
Security measures - having robust data security protocols in the event that, let's say, there's a hack in place, or a data breach, or somebody requests deletion of data. It’s incredibly important to have all security measures in place.
LBB> In your view, what’s the biggest misconception people have around the use of data in marketing?
Eric> For me, the biggest misconception is that data can replace human creativity.
Data provides valuable insights, but it cannot replicate the nuanced and imaginative aspect of human creativity. It will guide you somewhere, but ultimately it’s the creative leaders that are going to come up with the big ideas and transform insights into impact.
LBB> In terms of live issues in the field, what are the debates or developments that we should be paying attention to right now?
Eric> We're living in a time where technological developments are happening at such a fast pace that global privacy laws, or regulatory bodies, can barely keep up.
The change in global privacy laws, and the impact this has on data practices, is something we should be paying attention to. And Ethical AI is another one.
LBB> As technologies progress, and these systems become smarter and smarter, how can we ensure that AI-driven systems and the data analysis that they provide, and insights that they provide, are used ethically and responsibly?
Eric> Are we to say that a single company is responsible for this? And can we trust a single company to do it? We've seen this causing issues in the past, and so maybe there ought to be larger governing bodies that spread around countries to ensure fair, and ethical AI usage.
The last thing to keep an eye on is personalisation versus privacy. Now that we are able to have hyper-personalised marketing, we need to balance this with consumer privacy concerns. At the end of the day, we can only personalise marketing as much as we're able to collect data.
