Kevin Hillstrom – MineThatData EmailAnalytics is a publicly available dataset originally used in the MineThatData E‑Mail Analytics and Data Mining Challenge. It comprises data for 64,000 customers, each active within the past year, divided into three randomized groups: recipients of a Men’s merchandise email, a Women’s merchandise email, or no email (control group)
The dataset includes various historical customer attributes—such as months since last purchase (Recency), spending in the past year (History), spending categories (History_Segment), indicators for past purchases in men’s or women’s merchandise segments (Mens, Womens), geographic classification (Zip_Code), whether the customer is new (Newbie), and purchase channels (Channel)
Use case:
Email Campaign Effectiveness and Customer Segmentation Analytics
| Example Observation 𝐬 | 11 customer features captured before the test: • recency (days since last visit)• history (cumulative spend)• history_segment (binned spend)• last_update, zip_code, channel (mobile / web / multichannel)• newsletter (subscribed flag)• Demographics: segment, age, mens_category, womens_category, newbie |
| Example Action 𝐚 | Historical treatment assignment segment_email ∈ {0: No email, 1: Mens, 2: Womens}. (Uniform 1/3-1/3-1/3 randomised, hence good for OPE.) |
| Example Reward 𝐫 | spend (dollars in 2-week horizon). Alternate binary rewards are available (visit ∈ {0,1}, conversion ∈ {0,1}). |
| Example Transition | 64 000 independent customer trajectories ⇒ 64 000 (s,a,r,s′) tuples (no further steps because outcome measured once). |