Product
Profit-aware, margin-weighted bundle ranking with explainable math: more profit per bundle, not just more attach. The admin module turns order history into ranked bundles and opportunities across Shopify, BigCommerce, WooCommerce, Magento, and OroCommerce, with explainable support, confidence, and lift on every recommendation. The product does genuine behavioral market-basket mining and real catalog bundle SKUs, not impression-only widgets. Also agent-ready: the public API, the MCP server, and ACP endpoints make the same intelligence callable by AI agents and external apps.
Examples shown for illustration. Your results will vary based on catalog size, order volume, and margin structure.
New to market basket analysis? Start with Opportunities then move to Bundles & Kits.
Admin Intelligence & Bundle Creation
Every feature of the admin module, across Shopify, BigCommerce, WooCommerce, Magento, and OroCommerce, explained.
Analysis Jobs
- Run the pattern-finding algorithm against a chosen date window (30d, 60d, 90d).
- Choose between six selectable engines (SQL pairs, FP-Growth, seasonal cohort, return-aware, profit-aware HUI, and AI catalog), plus a co-occurrence fallback, with configurable thresholds.
- Verbose logging, warnings, and error tracking for every run.
Settings & Guardrails
- Set default time windows, min support, min confidence, and min lift thresholds.
- Configure engine-specific parameters (FP-Growth tree strategy, HUI utility floor).
- Guardrails protect against runaway jobs: max rules stored, min item frequency, max runtime.
Association Rules
- View the raw "if A then B" patterns the algorithm discovered.
- Filter by rule size (2-item, 3-item, 4-item) and sort by any metric.
- Every rule links to the Opportunities page for profit-aware scoring.
Opportunities
- Rules ranked and filtered by profit and revenue potential.
- Tabs for Top Opportunities, High-Margin, Emerging, and Underperforming Attach.
- One-click actions: view details, propose a bundle, or push cross-sell links.
Bundle Proposals
- Proposals include suggested name, SKU, pricing strategy, and estimated profit lift.
- Workflow: Pending Review → Approved → Create Draft → Enabled.
- Review, approve, or reject proposals with full audit trail.
Bundle Lifecycle
- Bundles are created as Draft first for review before going live.
- Manage pricing, images, and descriptions before enabling.
- Actions: open product, enable/disable, delete with full lifecycle tracking.
Performance
- KPI cards: bundle conversion rate, total revenue, total profit, and top performer.
- Per-bundle breakdown: sales, revenue, profit, attach rate, and trend vs. previous period.
- Identify which bundles drive results and which need repricing or replacement.
Reports
- Pre-built reports: Executive Summary, Bundle Pipeline, Margin Impact, Window Comparison.
- Preview reports inline or export as PDF, CSV, or JSON.
- Share results with stakeholders and track improvement over time.
Documentation
- Getting started guide with a 7-step setup checklist.
- Concept explanations for support, confidence, lift, and scoring.
- Engine comparisons, troubleshooting, and data privacy documentation.
The engine drives the cart, the checkout, and the experiment
The same co-purchase intelligence behind the admin and the storefront now powers the revenue moments after a shopper engages. Every surface below ships on Shopify, BigCommerce, WooCommerce, Magento, and OroCommerce, reuses the existing recommendation engine, and renders nothing when no candidates exist.
Cart drawer with AOV progress
- Opt-in slide-out cart with a free-shipping / AOV progress bar that nudges shoppers toward the next threshold.
- One to three in-cart recommendation upsells with one-click add.
- Upsells recompute as the cart changes, so the next-best add-on stays current.
Merchant A/B testing (Experiments)
- Run experiments on the recommendation surfaces with a deterministic control / treatment split.
- Results dashboard: impressions, conversions, conversion rate, AOV delta, and lift.
- Lift reported with a 95% confidence interval via a 2-proportion z-test. The Frequently Bought Together widget is wired end to end.
Post-purchase upsell
- After checkout, the thank-you / order-confirmation surface shows ranked recommended add-ons.
- One-click add, on the same co-purchase data as the rest of the engine.
- Ungated on the entry tier.
Broadened bundle formats
- Beyond real-SKU bundles, pick a bundle format per opportunity.
- Mix-and-match / build-a-box (BYOB), volume / quantity breaks (tiered), and BOGO / Buy X Get Y.
- Virtual-bundle render mode groups existing products with no duplicate SKU.
Public API + MCP for AI Agents
The same intelligence that powers the admin and the storefront is also a callable API. AI agents, Claude Desktop, Claude Code, the OpenAI Agent SDK, Cursor, Cline, can answer 'what goes with X?' with real co-purchase data from the merchant's actual order history.
Public REST API
`/api/v1/recommendations`Bearer-auth REST endpoint returning JSON. Mint per-shop API keys from the admin (shown once, stored as SHA-256 hash). Use it from your own app, a chatbot, an OpenAI assistant, or anything that speaks HTTP.
MCP Server
`@marketbasketanalysis/mcp` on npmOne-line install for Claude Desktop, Claude Code, Cursor, Cline, and the OpenAI Agent SDK. Exposes 19 agent-callable MCP tools (18 on non-Shopify platforms, since predict_reorder is Shopify-only) any MCP host can call when a shopper asks 'what goes with X?'
Want to wire this up? See the Public API quick start or the MCP server quick start.
Privacy-first by design
On the self-hosted modules (Magento, WooCommerce), your order data stays where you put it. AI features call your own provider account directly. Shopify and BigCommerce are MBA-operated hosted apps with per-store-isolated infrastructure, and OroCommerce is a thin hosted client that proxies mining to our hosted backend.
Local processing on self-hosted modules
On Magento and WooCommerce, mining, scoring, opportunities, and bundles run inside your own install. Order data never leaves your infrastructure. Shopify and BigCommerce run as MBA-hosted apps; OroCommerce is a thin hosted client that proxies mining to our hosted backend.
Bring-your-own LLM key
Plus AI features (ai_catalog, bundle copy) call Anthropic or OpenAI directly with your own key. On the hosted apps the key is stored encrypted at rest and decrypted per request; on the self-hosted modules it lives in your store. Either way your data goes to your provider account, not through ours. No surprise bills, no third-party data hop.
Hosted apps and the Hosted Engine
Shopify and BigCommerce are MBA-operated hosted apps with per-store-isolated infrastructure. OroCommerce is a thin hosted client that proxies mining to the same hosted backend, and the Hosted Engine add-on runs infrastructure on your behalf under a separately contracted DPA. Plus on Magento or WooCommerce never sends order data to MBA servers.
Ready to turn your order data into revenue?
Install on your platform in under 10 minutes. Or book a consulting call and we'll do the launch for you.