The two speakers from Oxyma, Christian Seidl (Junior Marketing Consultant) and Ruud van Luijk (Data Analyst) take the 10 attendees on an interesting journey. During the session all topics are discussed from both a technical and a marketing perspective. Next to a broad overview of what’s possible now and some practical cases, they also discuss two of the most important steps in the process of utilizing artificial intelligence: data collection/preparation and how to utilize the results of a prediction. Finally, the participants filled in the so called Machine Learning canvas to structure their analytical question.
Change is accelerating
We all feel that innovations and human progress are leaping forward faster and faster. Computing power is doubling every year. No wonder we think that we cannot keep up. There are hundreds of new ideas and innovations every day. Yes, AI has been a buzzword before, but this time it’s here to stay – making its way towards mainstream adoption. The conditions were not perfect before, but with faster hardware & cloud computing, more data & bigger datasets and advanced research & development we can finally put AI to use.
Applications of AI
How can companies use artificial intelligence to their advantage? The possibilities are endless. From self-driving trucks to doctors being supported by advanced image recognition algorithms that help to analyze MRI scans. Although this sounds very futuristic, most of these things are possible already – at least in first tests. During the session we discussed a few very practical applications of AI that are relevant for marketing. Describing the context of pictures, tagging or even creating images is all possible with the help of smart algorithms. Furthermore, you can predict the hotness of leads. Also, autonomous A/B testing is done by an AI campaign manager. These practices aren’t flawless yet: they require input and working hand in hand with a marketer, but they could potentially save a lot of time.
What do you need?
To be able to use AI in your organization you need to consider the 3 P’s. Process, platform and people. The process was discussed during a practical part of the workshop: by filling in the Machine Learning Canvas. This canvas starts at the value proposition and then covers the learning, predicting, and evaluating part of Machine Learning. By filling in this canvas, the participants were more accurately able to anticipate practical issues during such a project. The platform covers the right hardware and software (e.g. R, Python, SQL). It is of outmost importance that there is enough data and even more that the data quality is sufficient. Out of experience, 80% of resources and time in such Machine Learning projects should be dedicated on the part of data preparation, since this has the largest influence on the predictive strength of the model. Last but not least, you need the right people for the job. The algorithm can predict a certain outcome but human marketers are (still) needed to interpret the data and to come up with creative campaigns based on these insights.
Is AI the Holy Grail?
At the moment large companies that offer marketing tooling are definitely advocating this position. They praise AI as the all-knowing solution for every marketer. It’s true, AI is good at predictable tasks that can be automated and it can also predict human behavior and interests quite well. There are many practical examples where smart algorithms increased the ROI of marketing campaigns and increased efficiency.
However, you need a lot of data to be able to train algorithms. But even having plenty data is not good enough. You also need good quality data, often with labels (outcome X, outcome Y, etc.). By now, these models need supervision to yield the best results… and lucky for us marketers: we are still needed to interpret the results and create emotionally touching campaigns.
Lastly, it is not necessary to implement continuously improving, sophisticated models right away. It is possible to start with small, one-off experiments. Make a data export, train the model, predict the outcome, import & use the results for a campaign and hopefully you have shown enough business value to convince everybody to start with this exciting technology.
Curious as to how you can implement artificial intelligence in your marketing organization? For content-related questions please ask Ruud: firstname.lastname@example.org, and for more general questions please contact our Senior Data Strategist Remco: email@example.com.