I’m a data scientist not a public speaker so when the Keboola guys asked me to do a talk with them in London I was excited. The topic we chose was transactional data analysis mainly for two reasons – first, they can be used to solve so many business issues and secondly, they are everywhere.
It was a snow calamity in London but coming from -15°C we had in Prague felt like Spring was almost there. The event was hosted in Work.Life in Clerkenwell, which is cosy and welcome drinks make it even better. The audience was a mix of business owners, data scientists and marketeers from various industries.
After an initial intro about Keboola Connection I spoke about the importance of business imperative for successful data science. How we always start with the client’s business issue and how we translate the business issue into an analytical problem and only then look for data and features to use to solve it.
Because both business guys as well as data science practitioners are interested in real-life case studies I had prepared four cases – churn prediction, automated event detection, demand prediction and recommendation engine. Due to the lack of time we covered 3 of them. Recommendation engine being the one the audience put at the fourth spot.
The four cases were chosen because they were aiming to solve different business issue (retention, campaign efficiency, cost optimisation and x-sell/up-sell). They also represent four different analytical problems such as binary classification, anomaly detection, time series prediction and recommendations.
The group of people in the audience was reasonably small, which allowed for great interactions, questions and discussion about these problems. The questions ranging from technical ones (what features worked the best or which performance metric should be used for demand prediction) to the business ones such as what are the positive consequences of a churn model for the end customers.
All-in-all it was a great experience for me. I had to organise my thoughts and present for an hour to a well informed audience. I also hope that the attendees enjoyed the talk and found it useful for their jobs. And finally a big thank you to Keboola UK for making this happen!
Below are the slides I used. Credits for them goes to our friends at Odprezentuj.