24 February 2020


Impact, Explainability, and Resilience — Themes from Wisdom 2020

Is anyone else still recovering from Wisdom? Here at RapidMiner HQ, we’re getting back into the swing of things after all the fun from this year’s user conference. But the energy and insight that we took from the conference is going to power us for another year!

We loved engaging with our users, Unicorns and business track attendees across two jam packed days of presentations, trainings, friendly competitions, panels and more.

If you missed out on our conference this year—or just want to relive the experience—we’ve recapped some of the conference’s events below.

We chose these presentations because they highlight some of the prominent themes that were raised not only in these specific sessions, but which were woven into conversations and presentations throughout the conference.

Lighting fast business impact

This year we asked our users to address a problem plaguing cities and towns alike: missing, lost, and stolen packages. Using our data science platform, participants formed different teams to come up with their own predictive analysis models that predict which packages in the city of Boston were most likely to be stolen, and thus should be insured. Some teams took it a step farther and incorporated the cost of insurance into their model, making for a cost sensitive analysis.

One big shock to us as the organizers—although we’d allotted about four hours for the hackathon, all the teams finished an hour earlier than expected.  At the end of the hackathon, each team shared their model and explained how they arrived at their solution.

It was fascinating to hear the different ways that each team approached the problem, and how the RapidMiner platform was able to accommodate so many different routes to the same intended destination.

If you want to see the kinds of solutions that people came up with, including making them price sensitive, you can take a look at the solution from Martin Schmitz’s team, which he was kind enough to record after the fact:

Video Thumbnail









Resiliency over accuracy

Our co-founder Ingo Mierswa kicked off our second day of Wisdom with his keynote speech Deepfake is for losers. With 20 years of experience under his belt, Ingo feels strongly that the deepfake phenomenon is a waste of time for data scientists, noting that it will become increasingly difficult to differentiate between real and fake if the use of deepfake continues.

He’s developed his own manifesto for data science, to help make sure everyone is directing their time and efforts so as to make an actual, tangible difference in the world, rather than just creating more models that can’t be applied, thus perpetuating the Model Impact Disaster.

Some of Ingo’s guidelines for data science winners are:

The idea of resiliency was especially interesting to attendees—stay tuned for more from Ingo in the coming months on his data science manifesto, including how we can think about and measure resiliency.

Leading with predictions for enterprise AI

Our second keynote speaker was Vice President & Principal Analyst Mike Gualtieri from Forrester’s Research, speaking on Your future in enterprise AI is bright: You’ve built some models, now what?

Mike provided advice for those of us who intend to heavily implement AI in our organizations in the future, based on his many years of experience and research, including the need for data scientists to insist on comprehensive access to enterprise data, and demystifying machine learning for business people. It was an informative session that did a lot to alleviate concerns about implementing enterprise AI and painted an encouraging picture for the future.

One of the big takeaways here was to think about how you present AI projects to leadership. Rather than getting bogged down in the details of different kinds of models and algorithms, simply present the results of machine learning projects to leadership as “predictions”.

Not only are predictions something that leaders are used to dealing with from other fields like business intelligence but framing the problem as one related to prediction makes it easier to find uses for machine learning. Business leaders can easily think of things that they’d like to be able to predict ahead of time, which helps them see the value of AI right away.

Don’t wait for the perfect model

Our panel of experts, moderated by Michael Martin (Information Arts), included Brian Tvenstrup (Lindon Ventures), Cody Lougee (Vantage West), Martin Schmitz (RapidMiner), and Paul Simpson (Elliott Davis). They brought together their different perspectives as domain experts, data scientists, business users, and managers to discuss the different perspectives they have on data science projects.

As we saw demonstrated live in our hackathon, your professional background and training has a big impact on how you approach data science problems, and the best results tend to come when you can assemble a team that combines multiple different approaches.

The goal, naturally, should always be to work together towards a desired positive impact, rather than further contributing to the Model Impact Disaster. One of the major reasons people aren’t seeing the ROI they’d like on ML projects is that they hesitate to put models into production.

You’ll never have perfect data. Your model will never be perfect. The criteria for rolling out a machine learning project into production should simply be that it has some net positive impact on your business’s bottom line, even if the impact is small. So don’t wait for perfect model. If your model has impact, get it into production, and then iterate from there.

Making the intangible tangible

Heatherly Carlson and Joo Eng Lee-Patridge of Central Connecticut State University ran a fascinating user track session on sports analytics. The NFL Combine and draft processes have an impact on more than just your Fantasy Football draft selections.

They utilize data analytics to predict the future performance of players. Carlson and Lee-Partidge suggest that pre-draft metrics should look at more than just tangible variables, such as physical strength and agility, because that can change quickly and unexpectedly (and aren’t always the best predictors of success).

They raised the point that if all of the NFL teams attend the Combine, how is it that despite all teams receiving the same data set (player stats), some teams leave the draft with a more competitive advantage?

In exploring this question, Carlson and Lee-Patridge raised the idea of using intangible metrics to weigh draft prospects. Things like resilience to injuries, character risk, environment, and overall psychological makeup might have just as much to do with predicting the success of a quarterback as his physical strength.

This panel was an excellent demonstration of the practical uses for machine learning in every industry, highlighting the benefits of taking intangibles and turning them into something that can help you make better decisions.

AI and explainability

160over90 is an advertising agency using machine learning to help them target and developing campaigns, like this TV spot for Lifelight featuring Kristen Bell and Dax Shepard.

What really stood out about 160over90’s presentation is that they aren’t using complicated models and algorithms to support their work. Instead, they’re focusing on simple, easily explainable decision trees to help target ads. After walking decision-makers though a decision tree based on the ever-famous Titanic dataset, it’s easier for everyone to understand what the model is doing, which helps build buy-in from leadership.

This focus on explainability over complicated models further underscores the extent to which machine learning can have strong business impacts without introducing a lot of complications.


Our RapidMiner experts are always available to help guide users – but we have plenty of Unicorns in the RapidMiner community who know just as much as we do! This year we thought it would be fun to have a friendly competition to see who can really call themselves experts.

Team RapidMiner squared off against the Unicorns in a Family Feud trivia battle to test who knows the most about data science, machine learning, and artificial intelligence. Technically the winner was the Unicorn team, but the real winner was everyone who got to witness a real unicorn take the stage!

Until next time!

Thanks again to everyone who traveled to Boston to join us for the event! We’ll continue to be uploading video of the presentations from Wisdom over the coming weeks, so stay tuned for more Wisdom 2020 updates.

Still want more? Be sure to check out a variety of recorded presentations from our Wisdom 2020 user conference. Enjoy!

Related Resources