[Nature] Open e-commerce 1.0, five years of crowdsourced U.S. Amazon purchase histories with user demographics

[papy]

University researchers have published a free, open access data set of Amazon.com-based consumer behavior data, metrics, and analyses based on 5,000 participant purchases from 2018 to 2023. They report that ASIN-specific data are available.

• Article intro: Open e-commerce 1.0, five years of crowdsourced U.S. Amazon purchase histories with user demographics | Scientific Data
• Actual work: Berke, A., Calacci, D., Mahari, R. et al. Open e-commerce 1.0, five years of crowdsourced U.S. Amazon purchase histories with user demographics. Sci Data 11 , 491 (2024). Open e-commerce 1.0, five years of crowdsourced U.S. Amazon purchase histories with user demographics | Scientific Data
• Data set: GitHub - aberke/amazon-study
• Published: 13 May 2024

Some data of note, but much more available:

Spending trends

Top categories

[GGX]

Did they recruit participants at barnes and noble or something?

Book spend being #1 by a huge margin and more people buying books than a phone charger seems to indicate that this data is heavily biased.

[papy]

@GGX

Did they recruit participants at barnes and noble or something?

No.

Data collection

Survey participants were recruited via the online research platforms CloudResearch and Prolific.

We offered prescreen participants $0.35 for an estimated 1 minute survey. We offered participants $1.50 for the main survey with an estimated 4–7 minute completion time. Participants were paid whether or not they opted in to share their Amazon data. However, some participants received additional bonus payments of $0.05, $0.20, $0.50, based on the experimental survey design. More details are in the Supplementary Information (A). Data were collected in a series of batches between November 2022 and March 2023. We stopped collecting data on March 20, 2023 when Amazon took the Order History Reports page offline, which the survey tool depended on.

---

@GGX

Book spend being #1 by a huge margin and more people buying books than a phone charger seems to indicate that this data is heavily biased.

Just because you don’t like, didn’t expect, or don’t agree with the data doesn’t mean that it’s “heavily biased”.

The data did undersample those ages 18+ with only a high school diploma or less, as the researchers noted, who might be less likely to make book purchases. But that’s not “heavy bias”.

The sample has a slight gender bias with more females versus males, when compared to the U.S. population47. This is largely because females chose to share their Amazon data more often than males in the data collection process45. Our sample demonstrates an important age bias, under-representing older participants and over-representing younger participants48. The sample also under-represents higher-income households, while over-representing middle-income households49. Similarly, our sample over-represents individuals with a bachelor’s degree or greater level of education and under-represents those with a high school education or less49.