Choices between options

How strongly does price change which plan people choose?

The data. standard_conjoint.csv, the published sample file for Conjoint Analysis: 3,600 rows; columns respondent_id, task_id, price_usd, storage, support, chosen. 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_conjoint.csv")))
shown, chosen = {}, {}
for r in rows:
    p = int(r["price_usd"])
    shown[p] = shown.get(p, 0) + 1
    chosen[p] = chosen.get(p, 0) + int(r["chosen"])
tasks = len({r["task_id"] for r in rows})
print(len(rows), "options shown in", tasks, "choice tasks")
for p in sorted(shown):
    print(f"price {p}:", shown[p], "times shown,",
          f"{chosen[p] / shown[p]:.1%} picked")

3600 options shown in 1200 choice tasks
price 10: 847 times shown, 54.5% picked
price 20: 941 times shown, 40.7% picked
price 30: 894 times shown, 26.3% picked
price 40: 918 times shown, 13.1% picked

What it says. Each row is one option in a choice task, with one option picked per task. At the lowest price an option was picked 54.5% of the times it was shown; at the highest, 13.1%. Choice data has this shape: options grouped into tasks, not one row per person.

What it does not say. Storage and support vary across the same options, so this one-way table mixes their effect into price. The library analysis estimates a utility for every attribute level from the choices, which separates price from the rest.

Run Conjoint Analysis on this file or your own, or open its stored sample report.