The encyclopedia · Marketing & Brand · Marketing decision · 2000
Amazon charged different customers different DVD prices — for five days
A random discount test on 68 DVDs gave some customers 40% off and others full price. Forums noticed. Bezos called it 'dumb' and refunded 6,896 buyers.
Amazon · 2000-09
What happened
In September 2000 Amazon ran a five-day pricing test on 68 DVD titles, randomly showing different visitors discounts of 20% to 40%. Some customers saw the same DVD at $22 and others at $26. Amazon said the prices were assigned randomly, not by demographics or purchase history.
Customers on the DVD Talk forum compared prices and posted the discrepancies. Media coverage followed within days. The backlash was immediate: Amazon, which had built its brand on customer trust and low prices, was accused of price gouging.
Jeff Bezos issued a public apology, calling the experiment a mistake that 'created uncertainty for customers rather than simplifying their lives.' Amazon refunded the difference to all 6,896 affected customers — an average of $3.10 each — and pledged never to test prices based on customer data.
Why it happened
- A pricing test designed for economics was run on a customer base that had been promised low, transparent prices — the experiment contradicted the brand's core claim.
- Amazon assumed random assignment would go unnoticed; it underestimated how quickly online communities compare prices.
- No one modelled the reputational cost of a $3 discrepancy against the data value of the test.
The lesson
A pricing experiment on loyal customers is a trust withdrawal, not a test. If the data is worth less than the apology, don't run it.
Sources
spotted an error? The club wants to know.
More like this
Hopper hid a 'tip' and a VIP fee at checkout and paid the FTC $35M
Orbitz showed Mac users pricier hotels — and called it personalisation
PayPal bought Honey for $4B — then the coupon app stole creators' affiliate commissions
Somewhere, someone solved the problem this company failed at. 2nd Opinion →
Comments · 0
Sign in to join the comments.