Most frequent items
Fraction of distinct baskets containing each item.
Commerce analytics / reproducible pattern mining
Basket intelligence overview
Discover co-purchase relationships without confusing historical associations with causal sales lift.
DatasetSYNTHETIC DATA
Observed transactions
Highest-lift relationships
Rules meeting both thresholds, sorted by lift. Rare associations need careful review.
Association does not prove an offer will increase purchases. All example transactions are synthetic.
Rule quality landscape
Each point is a rule; horizontal position is confidence, vertical position is lift, and circle size represents support.
How to interpret a rule
Support: share of all baskets containing both sides.
Confidence: share of baskets with the antecedent that also contain the consequent.
Lift: confidence divided by baseline consequent frequency.
Leverage: excess joint frequency beyond statistical independence.
Confidence: share of baskets with the antecedent that also contain the consequent.
Lift: confidence divided by baseline consequent frequency.
Leverage: excess joint frequency beyond statistical independence.
Rule mining laboratory
Mine itemsets up to three items. Thresholds update every chart, table and export.
All measures are computed locally from distinct transactions.
| Rank | Association | Support | Confidence | Lift | Leverage | Evidence |
|---|
A rule A → B is directional as a conditional probability, but it does not establish time order, customer intent, or causation. High lift on small samples may be unstable.
Bundle discovery workbench
Select an anchor item to inspect historical add-on patterns, including absolute joint counts.
| Suggested add-on | Co-purchases | Conditional share | Baseline share | Lift |
|---|
This workbench includes every co-purchased add-on with observable support, not only rules passing the mining thresholds.
Segment intelligence
Compare item penetration and co-purchase behavior across groups. Segment samples may be unequal.
Popular items by segment
Evidence limitations
Comparing conditional shares across segments does not control for demographics, exposure, assortment, promotions or selection bias.
Small group sizes can produce unstable lift estimates. These groups are synthetic and do not describe real customers.
Scenario planner
Build a transparent hypothetical campaign using manually entered uptake and unit economics, not a causal estimate derived from the mining results.
Assumption sensitivity
Hypothetical net contribution for three uptake assumptions.
Historical context
A 5% incremental uptake entered here is only a scenario assumption. Observed confidence or lift cannot be substituted for incremental conversion or financial return without a controlled experiment.
Import transaction data
CSV must include transaction_id, item and segment; basket_value is optional. Each row represents one item in one basket.
Strict CSV validation, size limits and no external uploads. Imported data remain in the browser until you leave or reload this page. Export before closing.
Dataset diagnostics
Reproducible exports
Downloads reflect the currently selected segment and mining thresholds unless explicitly described otherwise.
Methodology
All metrics are derived from distinct transaction-item occurrences, so repeated rows do not inflate the support count.
Support = |A ∩ B| / N. Confidence = |A ∩ B| / |A|. Lift = confidence / (|B| / N). Leverage = support(A ∩ B) − support(A) × support(B).
Itemsets are exhaustively enumerated up to size 3. This is deliberately limited to manageable local datasets. No model is trained and no real-time market data are retrieved.
Support = |A ∩ B| / N. Confidence = |A ∩ B| / |A|. Lift = confidence / (|B| / N). Leverage = support(A ∩ B) − support(A) × support(B).
Itemsets are exhaustively enumerated up to size 3. This is deliberately limited to manageable local datasets. No model is trained and no real-time market data are retrieved.