The Analysis of Sensitivity of Demand for Food Products to Price and Non-price Factors (in Cooperation with Samokat.ru)

This academic study aims to get a directed causal graph, which demonstrates interconnections between variables, controlled by “Samokat” company (prices on products and minimal check), and demand on the products’ categories of this company in Saint-Petersburg. “Samokat” is a Russian online retailer, which sell food and household goods. The type of researches, when the causal structures of some processes are being studied on observational data is called “causal discovery” type. In addition to directions, we also try to get the effects of prices’ changes. Despite the algorithms, which was created to solve “causal discovery” tasks exists, they are not appropriate for our research due to violation of sufficiency assumption and lack of enough data. To achieve our goal, we apply the independent component analysis algorithm (ICA) to get the “additional variable”, which serves as a proxy of all unobservable variables. Such a proxy let us to measure the effects with regular linear regressions. According to the results, the most interconnected with demand on categories price variables are “Milk” category prices (21 effects), “Meat” (20 effects) category prices and minimal check (20 effects). The final graph can be used to measure the effects of company’s changes of prices. Moreover, it’s effects can be transformed into models’ “penalties” for increases of prices, which negatively influence on many products’ categories.