Marketing & Growth

Which media channel moves with revenue week to week?

Which channel moves with revenue is where media planning starts. This runs on the library's sample weekly media file.

The data. standard_mmm_lite.csv, the published sample file for Media Mix Model (Lite): 156 rows; columns Week Starting, Net Revenue, TV Spend, Paid Search Spend, Social Spend, Display Spend. 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_mmm_lite.csv")))
rev = [float(r["Net Revenue"]) for r in rows]
def corr(a, b):
    ma, mb = sum(a) / len(a), sum(b) / len(b)
    sab = sum((x - ma) * (y - mb) for x, y in zip(a, b))
    return sab / (sum((x - ma) ** 2 for x in a)
                  * sum((y - mb) ** 2 for y in b)) ** 0.5
print(len(rows), "weeks")
r_by = {}
for ch in ["TV Spend", "Paid Search Spend",
           "Social Spend", "Display Spend"]:
    r_by[ch] = corr([float(r[ch]) for r in rows], rev)
    print(ch, "vs revenue: r =", round(r_by[ch], 2))

156 weeks
TV Spend vs revenue: r = 0.44
Paid Search Spend vs revenue: r = 0.61
Social Spend vs revenue: r = 0.25
Display Spend vs revenue: r = 0.11

What it says. Of the four channels, Paid Search Spend tracks revenue most closely week to week (r = 0.61) and Display Spend least (r = 0.11). That is the first look a marketing mix model starts from, not its answer.

What it does not say. Search spend often rises because demand is already rising, so a correlation can run backwards. Media also works with a delay and wears out. The library analysis models carry-over and diminishing returns for all channels at once.

Run Media Mix Model (Lite) on this file or your own, or open its stored sample report.