Monthly Archives: February 2023

Analytics VI: Projects

Another year, another list of exciting projects (previous ones here, here, and here) from the course Analytics for Strategic Management, which I teach with my excellent colleague Chandler Johnson). In this course, students work on real data analysis projects for real companies – and here is a (rather disguised) list:

  • Avinor, the Norwegian Airport Authority, wants to predict TOBT (Target Off-Block Time) for Gardermoen Airport. TOBT is a measure of when the plane will leave the gate, and very important for planning access to runways and other congested areas of the airport.
  • Norsk Tipping AS wants to improve its marketing of certain products through predicting customers’ likeliness to adopt them (see this (Norwegian) article for a former, very successful project with this company)
  • An international company in the shipping supply business wants to predict prices for some of its products. A key issue here (as is often the case) is finding data from orders that were not accepted (and, hence, not registered anywhere). You cannot know what price a customer will accept unless you have access to cases where the price was too high.
  • A news agency wants to predict the uptake of its articles to prioritise its editorial resources. The get news articles from news agencies and other sources around the world, and need to know which of those to spend money on translating and editing for the Norwegian market.
  • A grocery wholesaler wants to predict demand for its products. in the grocery industry, most stock levels are determined either by having minimum levels or by going by what you bought last year with some adjustments. This group wants to see if they can improve on that.
  • An insurance company wants to predict churn for some of its products. This problem is common to any company running a subscription business – which is increasingly true for more and more companies.
  • An large business school wants to predict grades for large exams. Manually reading through thousands of exams is boring work – not to mention expensive – can machine learning in some form be used to automate some of the work?
  • A large engineering company wants to predict employee churn. Engineers and other specialists are difficult to find, and it is much cheaper to retain a good employee than to find a new one.
  • Brønnøysundsregistrene, a Norwegian register for, amongst other things, company annual accounts, wants to predict late submissions. If they can predict this, they can make efforts to follow up more carefully on those companies, rather than send out nagging reminders to everyone.

One problem we often have in these projects is difficulty in getting data. This is not the case this year. Whether this is a result of more companies saving more data, the students getting better at defining problems based on data they have, or just plain coincidence, remains to be seen. But it is a welcome development!