Quick Summary

This project explores basket composition and “splash effects” in retail shopping data. The goal is to determine which products act as “anchors” — items that, when purchased, are associated with larger baskets or higher-value co-purchases. By identifying these anchors, marketing recruitment dollars can be allocated more efficiently toward customers or product categories that generate higher ROI through expanded basket depth.

Methodology

1. Data Sourcing

• The dataset used is Instacart Market Basket Analysis from Kaggle, a Google-owned data science community with example datasets. The data sourced and utilized for this analysis is purely experimental and does not reflect real data.
• Transaction-level records include Transaction ID, Customer ID, Product ID/Category, Quantity, and Price.

2. Data Preparation

• Transform raw transactions into basket-level views (one row per transaction).
• Encode product presence using binary flags (1 = purchased, 0 = not purchased).
• Aggregate product sales frequencies to identify candidate anchors.

3. Anchor Selection

Anchors are chosen using a hybrid approach:

• Frequency-based: Top N most common items/categories.
• Value-based: High-revenue or high-margin products.
• Association-based: Items with strong co-occurrence patterns (high lift/support).

4. Basket Composition Analysis

For each anchor:

• Calculate average basket size (number of distinct items).
• Calculate average basket value (sum of prices).
• Compare across anchors to see which lead to larger or more valuable baskets.

5. Splash (Co-Occurrence) Analysis

• Build a co-occurrence matrix of anchor products vs. secondary products.
• Compute probability of co-purchase and lift ratios.
• Visualize with heatmaps or network graphs to show strongest splash effects.

Timing & Engagement

• Peak Day: Sunday (17.3% of orders)
• Lowest Day: Thursday (12.5% of orders)
• Peak Hour: 10 AM (8.5% of orders)

Recommendation: Schedule push notifications and promotions on Sunday mornings and reinforce mid-week traffic with Thursday specials.

Product & Assortment Strategy

Concentration: Top 20 products drive 10.3% of line items; top 100 capture 23.1% of total volume.

Recommendation: Feature hero SKUs in search banners and guarantee availability — these products are basket drivers. Maintain depth in high-frequency SKUs.

Cross-Sell & Bundling

High-Lift Pairs (from top_pairs_by_lift.csv):

• Blueberries ↔ Regular Sliced Bacon (Lift ≈ 57)
• Blueberries ↔ Jamaican Allspice (Lift ≈ 57)
• Blueberries ↔ Orange Liqueur (Lift ≈ 57)

Anchor Recommendations:

• By Lift: Yellow Onions → Original Lager (Lift ≈ 44)
• By Confidence: Yellow Onions → Banana (29%), Large Lemon (14.7%), Organic Baby Spinach (13.8%)

Recommendation: Bundle fresh produce (bananas, spinach, lemons) with staple onions in recipe kits; explore discovery bundles (e.g., produce + beverages) based on high-lift but niche pairs.

Category & Department Insights

• Department Penetration: Produce (75%), Dairy & Eggs (68%), Beverages (45%)
• Cross-Department Synergy: Alcohol + Other (Lift ≈ 3.1), Alcohol + Pets (Lift ≈ 2.5)
• Aisle Hotspots: Fresh Fruits (56%), Fresh Vegetables (44%), Packaged Vegetables & Fruits (37%)

Recommendation: Promote Produce + Dairy pairings (e.g., fruit + yogurt). Leverage alcohol cross-sell potential with pet or “other” categories for curated lifestyle bundles.

Loyalty & Reorder Programs

Replenishment: With a 59% reorder rate and top SKUs (bananas, spinach, strawberries) recurring heavily, there’s strong scope for auto-replenishment.

Recommendation: Offer subscription boxes or reorder reminders for high-frequency fresh items; trigger nudges around the 7–10 day median cycle.

Implementation Roadmap

Visualizations