Amazon Seller Support Contact Sequences with 548k Interactions

Amazon provides a dataset of 548,000 anonymized seller support contacts across 118 intents from 70,000 sellers sampled from recent years. The data includes de-identified seller IDs, perturbed inter-arrival times between contacts, and contact intent codes. It is released under the CDLA-Permissive-1.0 license.

Use Cases

• Predict the next contact intent for a seller based on their historical sequence of intents.
• Model the temporal patterns of seller support requests using the noisy inter-arrival times.
• Analyze the distribution and frequency of different support intents across the seller population.
• Train sequence models for next-timestamp prediction given a seller's contact history.

Strengths

• Contains 548,000 contacts, providing substantial volume for modeling.
• Includes 118 distinct intent categories, offering a wide range of support topics.
• Data is temporally ordered per seller, preserving sequence patterns for analysis.
• Inter-arrival times are perturbed to anonymize sellers while preserving temporal patterns.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the full dataset is unknown, which may limit suitability assessment.
• The inter-arrival times are described as 'noisy' and perturbed, which may affect precise temporal analysis.

Provenance

Source

Amazon

Collection Method

Sampled from recent years of Amazon seller support contacts via email, chat, or phone.

Time Range

Recent years (specific range not provided).

Freshness

Last update date is unknown; freshness unverified.

Geography

Global (inferred from Amazon's platform).