How Polywood is using AI to sharpen personalization, conversion

Title

How Polywood is using AI to sharpen personalization, conversion

Content

Polywood, a U.S. outdoor-furniture retailer operating on Shopify since a platform change roughly 18 months ago, has used Anthropic's Claude to help construct code bases and built an internal LLM-powered operating system that fuses MLS real-estate data, public weather feeds and historic purchase data to predict purchases, generate personas and personalise marketing — lifting conversion rate by 22% and average order value by 12%. 'We kept all of our developers. We didn't fire a single one. They are now twice as fast deploying new features, new capabilities,' said Ben Spiegel, Chief Digital Officer at Polywood, in an interview with Digital Commerce 360. On AI-generated marketing imagery he added: 'So not just studio shots, but also a lot of the imagery you see now is 100% AI, large-scale automated, which are my funnest projects.' Polywood set out to sharpen personalisation, conversion and developer productivity after switching ecommerce platforms. The retailer's three main prediction factors are weather, home sales and permits. Its internal LLM combines national weather data and Multiple Listing Service real-estate data with historical Polywood orders to predict what a customer is likely to buy — based on house age, purchase date and product category. Polywood switched to Shopify around 18 months ago, concurrently moving to '100% AI-based coding' and retaining its developer team as twice-as-fast feature deployers. The retailer has iterated on its internal LLM-based operating system since, adding chatbot-style natural-language queries against previous sales and trends, persona-based marketing creative and AI-rendered imagery across 150,000 SKUs. The interview with Spiegel was published August 12, 2026. Polywood uses Anthropic's Claude to help construct code bases and runs an internal LLM on its operating system that ingests MLS real-estate data and weather feeds alongside historical purchase data. The internal system exposes chatbot-style natural-language queries about previous sales and trends. Polywood reverse-correlated weather against orders and found that the seven-day forecast — not the day-of, three-day or 14-day forecast — is the leading indicator for a sale, then trained machine-learning models on those patterns. AI is also used to render product imagery at scale across 150,000 SKUs. After switching platforms and adopting AI-based coding, conversion rate rose 22%. AI-personalisation lifted AOV by 12% — material given Polywood's average purchase price of around $1,600. AI-driven ad copy and personalised landing pages decreased cost per click and contributed to the conversion lift. The retailer now generates AI imagery for 150,000 SKUs in multiple colours. Polywood plans to extend persona-based marketing, continue building toward lifestyle imagery rather than just studio shots, and use its internal LLM to drive attribution modelling, ROAS targets, AOV optimisation and personalisation across the retailer.