Statistics Quest

Weekly store sales

Weekly sales for the 40 stores of a fictional home-goods chain, Lumen Bay Stores, over two years (104 weeks). Each row is one store-week with the store's region and size, a holiday-week flag, a promotion flag, footfall and sales in SAR. Sales follow a gentle upward trend, annual seasonality, holiday peaks and a real promotion lift, so the data support ANOVA by region, regression with dummies and time-series decomposition.

Open in the labDownload CSV
Rows
4,160
Columns
10
One row is
one store-week
Version
v1

Data dictionary

ColumnTypeUnitSummaryDescription
store_id
Store
idStore identifier (ST01–ST40).
region
Region
category4 levels: Central, Eastern, Western, SouthernSales region.
store_size
Store size
ordinal3 levels: Small, Medium, LargeSize class of the store.
floor_area_m2
Floor area
integerm²mean 1564, median 1510, SD 660.2, range 690–2990Sales floor area.
week
Week number
integermean 52.5, median 52.5, SD 30.02, range 1–104Week index from 1 (first week of 2023) to 104.
week_start
Week start
dateFirst day (Sunday) of the week.
holiday
Holiday week
boolean2 levelsTrue when the week contains a public/festival holiday period.
promotion
Promotion
boolean2 levelsTrue when the store ran a chain-wide discount promotion that week.
footfall
Footfall
integervisitorsmean 3417, median 3109, SD 1412, range 1088–11150Number of visitors counted at the entrance.
sales_sar
Weekly sales
numberSARmean 323,020, median 293,911, SD 128,382, range 114,700–966,480Net sales for the week.

What it is good for

  • Compare mean weekly sales across regions and store sizes with one-way ANOVA.
  • Estimate the promotion lift with regression while controlling for store, season and holidays.
  • Decompose a store's sales series into trend, seasonality and noise.
  • Explain why store-weeks from the same store are not independent observations.

Methods: one-way ANOVA, two-way ANOVA, multiple regression, dummy variables, time series decomposition, moving averages.

Data quality notes

  • Observations are clustered: 104 weeks per store. Treating all 4 160 rows as independent overstates precision.
  • Promotions are more likely in holiday weeks, so a naive promoted-vs-not comparison mixes the promotion lift with the holiday effect.
  • Holiday weeks: 15, 16, 26, 38, 51, 66, 67, 77, 90, 103 (lunar-calendar holidays fall about 11 days earlier in the second year).
  • No missing values.

Provenance and licence

Each store gets a base weekly sales level from its size (small 180k, medium 300k, large 480k SAR), a region factor and a log-normal store effect (sd 0.15). Weekly sales = base × trend (+0.08% per week) × annual seasonality (±7% sine wave plus a summer dip) × 1.35 in holiday weeks × 1.12 when the store runs a promotion (≈15% of store-weeks, more likely around holidays) × log-normal noise (sd 0.06). Footfall = sales ÷ average spend per visitor (≈95 SAR, varying by store) with Poisson-like noise.

Synthetic data · CC0-1.0 (synthetic data generated by Statistics Quest) · generated deterministically (seed 20241003). It does not describe real people or organisations; any resemblance is coincidental.

Preview

Dataset

Weekly store sales

4,160 rows · one row = one store-week · synthetic data

About this datasetOpen in the lab
StoreRegionStore sizeFloor area (m²)Week numberWeek startHoliday weekPromotionFootfall (visitors)Weekly sales (SAR)
ST01CentralSmall95412023-01-01falsetrue1986214951.9
ST01CentralSmall95422023-01-08falsefalse1811201838.41
ST01CentralSmall95432023-01-15falsetrue2045217021.33
ST01CentralSmall95442023-01-22falsetrue1947212706.67
ST01CentralSmall95452023-01-29falsefalse1798193921.23
ST01CentralSmall95462023-02-05falsefalse1894201477.48
ST01CentralSmall95472023-02-12falsefalse1830197267.45
ST01CentralSmall95482023-02-19falsefalse2001206280.41
ST01CentralSmall95492023-02-26falsefalse2025211217.87
ST01CentralSmall954102023-03-05falsefalse1853201213.28
ST01CentralSmall954112023-03-12falsefalse2113206739.23
ST01CentralSmall954122023-03-19falsefalse1865199818.51

Used in