Retail performance review: two years of Lumen Bay Stores

Lumen Bay Stores (a fictional home-goods chain with 40 stores in four regions) is preparing its two-year review for the board. You are the lead analyst. Three decisions hang on your work:

1. Growth. The CEO will tell the board how fast the business is really growing. Is 2024 better than 2023, and by how much?
2. Formats. The property director has budget for new stores in 2025 and must choose between the Small, Medium and Large formats. Which format earns most from its floor space — and is the evidence strong enough to bet on?
3. Promotions. The chief financial officer funds a chain-wide discount promotion that individual stores can opt into in any week. In 2024 it ran in 359 store-weeks. For this review, assume finance's scenario figures: each promoted store-week carries about SAR 22,000 of costs not already netted out of sales, and the gross margin is 38%. (These are planning assumptions for a fictional chain, not industry benchmarks.) Should the programme continue in its current form?

Stakeholders. The CEO and board want a clear story; the CFO wants money and risk; regional directors want to know whether their region is different; the property director wants a format recommendation that will survive scrutiny.

Constraints. Stores chose when to promote, so this is observational data, not an experiment. Holidays follow the lunar calendar and move about 11 days earlier each year. The board paper is limited to one page plus an appendix.

The data. retail_weekly has 4,160 rows — one per store-week for 40 stores × 104 weeks (2023–2024) — with region, store size, floor area, holiday and promotion flags, footfall and net sales in SAR. It is synthetic: generated to behave like real retail data (trend, seasonality, holiday peaks, a promotion effect), but it describes no real company. Where the lab cannot aggregate the data for you, the tasks give you store-level tables computed from it.