PoolPartsToGo — Data & AI Foundation for Ecommerce Growth
Single source of truth, actionable analytics, and fast delivery to transform operations and customer experience.
Team
| Role | Name |
|---|---|
| Project / Account | Amber Lin |
| Engineering / AI | Miguel de Veyra |
| Data / Analysis | Robert Tseng |
| Executive Sponsor | Uttam Kumaran |
| Client Lead | Daniel Schonfeld |
| Client (Operations / Product) | Ben [Cohen] |
| Client (Marketing / Dashboards) | Kim Todaro |
Agenda
- Context — Who PoolPartsToGo is and why data mattered
- Challenge — Fragmented data, no single source of truth, operational opacity
- Solution — Data & analytics (UPS modeling, forecasting, negotiation) and AI (pool Q&A chatbot)
- Results — Quantified savings, delivery speed, handoff
- Impact — Pilot to rollout, workflow integration, relevance for integrated ecommerce
Client Overview
PoolPartsToGo (PP2G) is a leading e-commerce company in pool parts (pumps, covers, ladders, equipment) selling across the US via Amazon, Shopify, and other channels. They run DTC, retail, and wholesale with a large product catalog and multi-channel sales. Leadership treats data and analytics as central to how they run the business and compensate teams, so a reliable, unified view of operations was essential.
The Problem
- Contract and cost opacity — No clear view of UPS contract terms or how they were charged; impossible to forecast or optimize shipping spend.
- Multi-channel complexity — Sales across Amazon, Shopify, and others with hundreds of SKUs and scattered data; no single place to see performance.
- Customer Q&A and support load — Pool owners relied on unstandardized sources (Facebook, Reddit); one-size-fits-all advice fell short and created noise.
- Legacy tools and silos — Data in separate systems (Zendesk, ad platforms, marketplaces); executives lacked a single, accurate, timely view for decisions.
Challenge: No single source of truth for shipping or logistics; slow, reactive decisions; no branded, standardized way to serve pool owners; difficulty scaling from reporting to proactive analysis (pricing, bundling, SKU optimization).
The Solution
Brainforge acted as an extension of PP2G’s team in two areas: data & analytics (UPS work as flagship) and AI / digital experience (customer-facing pool Q&A and roadmap to DIY and product recommendations).
Data & analytics
- UPS contract modeling and forecasting — Mapped contract to a rule-based (SQL) model that aligned any order to package price and actual invoices; used to forecast shipping costs by volume, growth, and new markets.
- Negotiation support — With a clear forecast (hundreds of thousands in spend), Brainforge supported and joined conversations with UPS and FedEx as the data authority on the client’s shipments.
- Dashboards and daily operations — Daily KPI and other dashboards used by the client (e.g., Kim), with regular syncs for accuracy and maintenance.
- SKU, margin, and pricing analysis — Data cleanup and analysis to support pricing, bundling, and top-SKU performance (e.g., variable-speed pump line).
AI / digital experience
- Customer-facing pool Q&A chatbot (pptg.ai) — Branded web app for general pool questions (chemistry, “why is my pool green?”, nearest store, product advice). Uses curated scraped content and Serp API; supports image upload and analysis; conversations logged to S3 and Google Sheets. Core chatbot built in ~2 days; full POC (scraping, UI, pipelines) in ~2 weeks with handoff and internal QA.
- Roadmap — DIY flow (step-by-step installation) and plans for image-rich demos and product recommendations.
Platform (summary) — Snowflake warehouse; dbt for modeling; Fivetran and direct integrations (Shopify, Amazon, Walmart, Zendesk, Klaviyo, paid ads, Primus, Unis, Shipstation, LTL); Rill Data for dashboards (6-hour refresh); Evidence.dev and Python/Prophet for reporting and forecasting.
Technologies Deployed
- Data & analytics: SQL/rule-based contract logic; Python/Prophet forecasting; dbt (dbt-snowflake) in Snowflake; Rill Data (Daily KPI, Shipments, Orders, marketing, returns, inventory, Zendesk, Kim’s Weekly Report); Fivetran and direct source connectors.
- Chatbot: MERN stack; agent with scraped FAQs; Serp API; image analysis (vision); S3 and Google Sheets for conversation logging and analytics.
Results and Impact
Results
- Shipping — Contract model and forecast enabled negotiation with UPS and FedEx; ~80–90% discounts on rates; savings in the hundreds of thousands of dollars (Brainforge involved from contract understanding through forecast and negotiation).
- Speed — AI POC (pool Q&A with image support) delivered in ~2 weeks with QA and handoff.
- Operational use — Dashboards back in active use; client uses the system for daily decisions; regular syncs (e.g., with Kim) keep dashboards aligned.
Impact
- Pilot to rollout — Data work became ongoing infrastructure (model, forecasts, dashboards) and negotiation execution; AI POC became deployable app (pptg.ai) with path to broader rollout.
- Workflow integration — Data and dashboards used for daily KPIs, pricing, campaigns; chatbot positioned for domain deployment and top-of-funnel/support.
- Replication — Same pattern (unify data, model costs, forecast, negotiate) applies to other carriers or regions; chatbot pattern (curated knowledge + search + image) extends to DIY and product recommendations.
Closing
PoolPartsToGo shows how a single source of truth and actionable analytics drive direct dollars: from no visibility into shipping contracts to modeled, forecasted view and negotiated savings in the hundreds of thousands, and from scattered customer Q&A to a branded, fast-to-ship AI assistant. For e-commerce leaders, the story reinforces that integrated data systems, clear ownership of data-to-dollars projects, and fast, documented delivery with handoff are the foundation for scaling operations and customer experience.
Sources: Brainforge vault transcripts (PP2G/PPTG/poolpartstogo); PP2G code repository (brainforge-ai/poolpartstogo). Metrics and quotes from transcripts; approximate figures stated as such.