Skip to content
Contact UsStart Building

Retail & E-commerce

Online purchase intent

Estimate purchase intent from browsing behavior to prioritize customer engagement.
AI-generated representative illustration of unbranded corrugated shipping parcel with neatly folded apparel and a closed slim laptop.
AI-generated task illustration. Representative scene; not a customer site or benchmark observation.
From known examples to a prediction

Available inputs

  • Page visits and browsing duration
  • Bounce rates and session context

Known examples

12,330 recorded examples

Prediction target

Whether the session results in a purchase
0.33 pphigher ROC AUC

vs. XGBoost (tuned ensemble), the strongest conventional result in this benchmark.

Online purchase intentROC AUC · Higher is better

12,330 examples · 17 inputs · 9 evaluation splits

01
  1. Causilo
    0.9389
  2. XGBoost
    0.9355
  3. LightGBM
    0.9355
  4. CatBoost
    0.9343
  5. Random forest
    0.9316
  6. Extra trees
    0.9311
  7. K-nearest neighbors
    0.9219
  8. Logistic regression
    0.9176

The task in practice

The public dataset contains browsing sessions with page visits, browsing duration, bounce rates and session context. The target records whether the session ended in a purchase.

  1. Connect sessions with purchases. Use recorded browsing sessions and their purchase outcomes as context.
  2. Describe another visit. Supply the same session analytics without the purchase outcome.
  3. Evaluate engagement priorities. Check the intent scores on independent sessions before testing how they inform customer engagement.

Scope. Session-level benchmark results do not establish that an intervention increases conversion.

Dataset, sources and evaluation

Online purchase intent dataset. Original source and curated OpenML task linked below. Original dataset source.

The TabArena-curated OpenML version (46947) supplies the dataset and recorded target, Revenue. Comparison scores come from the retained results_per_split_official.csv evaluation snapshot. These pages describe the task and compare the recorded means; source records are not republished here.

ROC AUC measures how well the scores rank the recorded positive outcomes above negative ones. A higher score is better. The change is reported in percentage points, rather than relative error reduction.

Causilo uses its default configuration. Each conventional family uses its strongest reported mean across default, tuned and tuned-ensemble settings. The chart compares reported means, without a claim of statistical significance, customer deployment or operational savings.

ROC AUC difference in percentage points = (Causilo mean − strongest conventional mean) × 100. ROC AUC is a ranking metric, not a percentage of correct predictions.

How to read these comparisons · All configurations and original values.