How Polywood is using AI to sharpen personalization, conversion

Outdoor furniture retailer Polywood has been using artificial intelligence (AI) and machine learning to unpack customer trends and strengthen its personalization, according to Ben Spiegel, its chief digital officer.

The retailer uses Anthropic’s Claude to help construct its code bases, Spiegel told Digital Commerce 360. He said that it also has developed its own large language model (LLM) that it has built to answer specific inquiries about consumer data and trends, which helps Polywood with retargeting.

In addition, Polywood changed ecommerce platforms about a year and a half ago, now operating on Shopify. That switch came with a 22% increase in conversion rate as Polywood “switched to 100% AI-based coding,” he said.

“But we kept all of our developers,” Spiegel explained. “We didn’t fire a single one. They are now twice as fast deploying new features, new capabilities.”

He said there has been a trend of AI replacing developers. However, Polywood is taking the approach that its developers “can do so much more” in developing ecommerce features when using AI as a tool rather than a replacement.

Reverse-correlating Polywood sales using AI

Spiegel said Polywood uses three main factors to “predict ecommerce”: weather, home sales and permits. To do that, it uses national, publicly available weather data that it ingests into its own operating system, which features its LLM. Polywood also bought multiple listing service (MLS) real estate data.

“And we joined that together with the weather data, and then we ran it against our existing old purchases to establish, based on your house, what are you likely to buy? And then we put in home purchase dates to be able to see: How long after you move into a house do you buy?” he shared.

It also used that mix of data to reverse correlate the weather’s impact on its orders.  Using historical weather and customer data, Polywood found that the day of an order, three-day forecasts and 14-day forecasts proved less indicative of whether a customer would buy from the retailer. Instead, Polywood found that the seven-day forecast is its leading indicator for a sale.

“It’s machine learning,” Spiegel said. “I hate to overuse the word AI, but we trained our machine learning models on those weather patterns to then be able to test it against the recent data and now predict the right time to send a catalog to serve an ad based on regionalized weather.”

Where the AI comes in, he said, is through Polywood’s internal operating system. His team’s LLM-based operating system has chatbot functionality. The team can ask any question it wants about previous sales and trends. 

Using AI to create customer personas

Polywood has used those three main predictors and its AI to develop customer personas. It then tests marketing messages based on those personas, which are more based on home types than personalities, he said.

He gave an example that consumers who buy a new house tend to buy outdoor furniture within the first six months. However, those who buy houses with pools tend to buy outdoor furniture in the first month, according to Polywood data.

Such data, in addition to millions of historic Polywood orders, allows the retailer to identify repeat order patterns, Spiegel explained.

Spiegel said the personas Polywood creates help the retailer to:

1. determine which creative to show.
2. drive the retailer’s own attribution models.
3. drive its parameters for both personas and targeting.

He noted that based on house personas, AI helps Polywood to determine strategy, return on ad spend (ROAS) targets, average order value (AOV) optimization and personalization.

Developing marketing creatives with AI

Polywood is also using AI to create marketing copy and images. It personalizes that content based on the ads consumers click on to arrive at the retailer’s website.

If a consumer searches for white outdoor furniture, Spiegel said, Polywood will use AI to display white outdoor furniture for that consumer in ads. It will then display white products to match on the landing page where the consumer arrives on Polywood’s site. 

That AI-powered personalization has helped Polywood to increase its AOV by 12%, “which at our price point is a lot,” he pointed out. He shared that Polywood’s average purchase price is around $1,600, which means its consumers are largely not checking out spontaneously.

Polywood’s AI-powered personalization also has helped to increase the retailer’s conversion rate by 22%, according to Spiegel. In addition, Polywood has used AI to generate ad copy, which has decreased its cost per click and contributed to increasing its conversion rate.

“We don’t always know which one to attribute which term,” he stated. “Was the ad better or was the landing page experience better? But we saw improvements in both of them.”

And Polywood has “evolved into lifestyle” imagery, Spiegel assessed. Polywood feeds its product models into its AI, which then renders images. The retailer has 150,000 different SKUs in different colors, and the imagery for them is all AI-driven now.

“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,” he said

Do you rank in our databases? 

Submit your data and we’ll see where you fit in our next ranking update.

Sign up

Stay on top of the latest developments in the online retail industry. Sign up for a complimentary subscription to Digital Commerce 360 Retail News. Follow us on LinkedIn, TikTok, X (formerly Twitter), Facebook and YouTube. Be the first to know when Digital Commerce 360 publishes news content.

The post How Polywood is using AI to sharpen personalization, conversion appeared first on Digital Commerce 360.