Retail Data Science Interview: Analyzing Promotion Impact
One frequent challenge in retail is understanding whether a sales spike during a promotion genuinely grows the business—or just shifts demand between products. This article dives into a real-world data science interview scenario from the retail sector, focused on quantifying promotional lift, cannibalization, and planning for future demand. We'll break down the solution step by step, explain all involved concepts, and provide illustrative charts and code snippets to help you ace similar interview questions—and build practical solutions in your own work.
A CPG (Consumer Packaged Goods) brand recently launched a two-week promotion for a new cereal flavor in 200 stores. The promoted item’s sales spiked by 60%, but overall category revenue barely changed. The revenue team suspects "cannibalization"—where the new flavor stole sales from existing ones. Meanwhile, they also want to understand any "halo effects" on complementary products like milk. The supply chain team is cautious, having previously over-forecasted and ended up with excess inventory. They want a practical, repeatable approach before the next promotional wave.
The first step is to figure out how much of the sales spike was truly incremental. Did the promo actually bring in new buyers or simply shift sales from other products or time periods?
