Weekly advertising and sales

Three years (156 weeks) of weekly marketing data for a fictional packaged-juice brand, Oasis Grove Drinks: spend on TV, digital and radio (thousand SAR), average shelf price, season, whether the main competitor ran a promotion, and units sold. TV has diminishing returns, digital spend tends to rise with TV campaigns, and there is a gentle upward trend — useful for regression with interpretation and its limits (observational data, confounding, collinearity, nonlinearity).

Open in the lab Download CSV

Rows: 156
Columns: 9
One row is: one week
Version: v1

Data dictionary

Column	Type	Unit	Summary	Description
`week` Week	integer		mean 78.5, median 78.5, SD 45.18, range 1–156	Week index 1–156.
`week_start` Week start	date			First day of the week.
`season` Season	category		4 levels: Winter, Spring, Summer, Autumn	Season of the week.
`tv_spend_ksar` TV spend	number	thousand SAR	mean 41.18, median 35.95, SD 42.33, range 0–123.9	TV advertising spend.
`digital_spend_ksar` Digital spend	number	thousand SAR	mean 14.2, median 13.9, SD 7.548, range 1–32.2	Digital (search + social) advertising spend.
`radio_spend_ksar` Radio spend	number	thousand SAR	mean 6.922, median 5.7, SD 7.867, range 0–31.6	Radio advertising spend.
`avg_price_sar` Average price	number	SAR	mean 24.55, median 24.68, SD 1.318, range 20.56–28.07	Average shelf price per 1-litre pack.
`competitor_promo` Competitor promotion	boolean		2 levels	True if the main competitor ran a price promotion that week.
`sales_units` Units sold	integer	packs	mean 7695, median 7723, SD 1499, range 4367–11250	Packs sold to consumers (retail scanner data).

What it is good for

Fit simple and multiple regressions of sales on spend and interpret coefficients in business terms.
Detect diminishing returns to TV with a residual plot and fix it with a log transformation.
Explain why digital's simple-regression slope overstates its effect (it moves with TV).
State the limits of regression on observational marketing data (confounding, causality, extrapolation).

Methods: simple regression, multiple regression, log transformation, residual plots, multicollinearity, dummy variables, prediction intervals.

Data quality notes

TV spend is zero in many weeks (between campaign flights).
TV and digital spend are positively correlated, which inflates standard errors and confounds simple comparisons.
Sales respond to TV with diminishing returns (logarithmic), so a straight-line TV term misfits.
Weeks are consecutive; residuals may show mild autocorrelation through the trend and seasons.
No missing values.

Provenance and licence

TV runs in campaign 'flights' (on for 4–8 weeks, then off), with spend 30–120k SAR when on. Digital spend = 8 + 0.15 × TV + noise (so they are correlated); radio is independent. Units = 5 200 + 1 400·ln(1 + TV/20) + 38·digital + 12·radio − 180·(price − 25) − 600·competitor promo + season effect (summer +900, spring +300, winter −400) + 6 per week of trend + Normal(0, 350).

Synthetic data · CC0-1.0 (synthetic data generated by Statistics Quest) · generated deterministically (seed 20241017). It does not describe real people or organisations; any resemblance is coincidental.

Preview

Dataset

Weekly advertising and sales

156 rows · one row = one week · synthetic data

About this dataset Open in the lab

Week	Week start	Season	TV spend (thousand SAR)	Digital spend (thousand SAR)	Radio spend (thousand SAR)	Average price (SAR)	Competitor promotion	Units sold (packs)
1	2022-01-02	Winter	0	5.8	12.4	23.85	true	4367
2	2022-01-09	Winter	0	9.8	19.2	26.05	false	5380
3	2022-01-16	Winter	75.3	16	0	25.35	false	7627
4	2022-01-23	Winter	61.3	21.7	0	24.7	false	7919
5	2022-01-30	Winter	62.6	23.1	11.6	25.97	true	6688
6	2022-02-06	Winter	58.6	24.7	6.7	23.47	false	7884
7	2022-02-13	Winter	66	22.5	6.1	25.83	false	7739
8	2022-02-20	Winter	0	16.1	11.6	24.65	false	5879
9	2022-02-27	Winter	0	7.7	7.7	26.4	false	4796
10	2022-03-06	Spring	0	14.2	0	24.38	false	6707
11	2022-03-13	Spring	0	11.3	0	25.14	false	5775
12	2022-03-20	Spring	60	16.5	0	26.77	false	7577

Used in

Scatter plots · Visualising and Exploring Data
Covariance and correlation · Correlation and Regression
Simple linear regression · Correlation and Regression
R² and residual standard error · Correlation and Regression
Residual analysis · Correlation and Regression
Inference for regression · Correlation and Regression
Multicollinearity · Correlation and Regression
Transformations · Correlation and Regression
Overfitting and validation · Correlation and Regression