At the Intellect Foundation’s Summer School, participants traditionally work with real business cases. The team led by VMK MSU student Ruslan Voit addressed a task for Ingosstrakh: within the case, they had to optimize marketing campaigns, consider the constraints, and identify a solution that would enable the company to maximize profits under the specified parameters. Ruslan spoke about finding a balance between the mathematical model and business constraints, as well as about what helped them bring the task to a successful conclusion.
– Ruslan, what project task did your team address at the Summer School, and in your opinion, what was its main challenge?
– At the Summer School, my team worked on a task from Ingosstrakh to optimize marketing offers. We had to figure out which clients should be offered a specific product and through which channel, so that the company could earn more without violating the business rules. The biggest difficulty wasn’t the math itself, but the initial data: the constraints were fragmented and incomplete. We first had to learn how to understand this data.
– How did you approach solving the case: where did you start, and what hypotheses or ideas did the team discuss?
– We started with methods that didn’t violate any of the constraints. However, this approach took too long to compute and ran into millions of data rows. Over time, we decided to allow the model to violate the constraints but to penalize it for doing so. This not only sped up its operation but also helped show the business which constraints were most hindering higher profits.
– Which knowledge or skills turned out to be the most useful while working on the case?
– What helped me the most was the ability to communicate with the client, not just with code. After a couple of meetings, it became clear that the real issue wasn’t a lack of an optimizer, but the time analysts spent translating the initial constraints from business language into a program. We shifted our focus in that direction and came up with a solution that can also be applied to other domains within the company.
– What solution did you ultimately arrive at, and how, in your view, could it be further developed?
– Ultimately, we ended up with a model that automatically interprets the constraints, finds the optimal set of offers, and suggests to the business which rules should be adjusted to increase profits. There are two ways to develop it further: either to make the forecasting smarter by taking into account the client’s history and the real effect of each contact, or to turn the tool into a full‑fledged product with a user‑friendly interface for managing constraints and optimization.
– What key skills did you manage to improve while working on the case?
– First and foremost, during the school, I improved my ability to finish tasks within tight deadlines and to organize team processes effectively. Networking with industry representatives was also valuable, as it opens up opportunities for future collaboration.
