What's actually driving this?

How do weekly sales move with ad spend?

The data. standard_linear_regression_v2.csv, the published sample file for Linear regression: 240 rows; columns ad_spend_k, region, sales_k, unit_price, week. Sample data, not a customer's.

The calculation. Python, standard library only. This exact code ran against the file when this page was built, and printed the output below it.

import csv
rows = list(csv.DictReader(open("standard_linear_regression_v2.csv")))
x = [float(r["ad_spend_k"]) for r in rows]
y = [float(r["sales_k"]) for r in rows]
n = len(x)
mx, my = sum(x) / n, sum(y) / n
sxy = sum((a - mx) * (b - my) for a, b in zip(x, y))
sxx = sum((a - mx) ** 2 for a in x)
syy = sum((b - my) ** 2 for b in y)
slope, r = sxy / sxx, sxy / (sxx * syy) ** 0.5
print(n, "rows")
print("sales_k =", round(my - slope * mx, 1), "+", round(slope, 2),
      "x ad_spend_k")
print("correlation r =", round(r, 2), " r squared =", round(r * r, 2))

240 rows
sales_k = 58.3 + 6.57 x ad_spend_k
correlation r = 0.93  r squared = 0.87

What it says. In this file each extra 1k of ad spend goes with about 6.57k more sales, and ad spend alone lines up with 87% of the variation in sales across these 240 rows, one per region and week (r = 0.93).

What it does not say. This is an association, not the effect of spending more: weeks with more spend may also be busier weeks, and the four regions are pooled here. One predictor is used here; the library analysis takes up to five drivers, such as region and price, reports an interval for each coefficient, and shows the residuals.

Run Linear regression on this file or your own, or open its stored sample report.