Latest / The DTC Marketing Podcast with Fexingo: Direct-to-Consumer Brands, Shopify, and E-Commerce Growth / How DTC Brands Use AI Chatbots to Reduce Return Rates
Transcript
- Lucas: Returns are the silent profit killer in DTC. In apparel and footwear, return rates can hit 40 percent or more. That's not just lost revenue — it's shipping costs, restocking labor, and the environmental toll of all that back and forth. Luna: And the customer who returns is often lost to the brand, too. There's this stat from Narvar that says 79 percent of shoppers won't buy again from a brand that made the return process painful. Lucas: Exactly. So the obvious question is — how do you prevent the return in the first place? And the answer a growing number of DTC brands are landing on is ai powered shopping assistants and chatbots that help customers get the right product the first time. Lucas: I want to ground this in a specific example. Warby Parker — the online eyewear brand — rolled out a virtual try-on feature using augmented reality. You hold your phone up, and it shows you how the glasses look on your actual face. The result? Their return rate dropped by about 11 percent. Luna: That's huge for a category where fit is everything. Glasses are deeply personal — the wrong bridge width or temple length and it's going back. Lucas: And Warby Parker is the poster child, but there are now dozens of brands doing this with AI chatbots that ask sizing questions, analyze past purchases, and even use computer vision to estimate your body measurements from photos. Luna: I've seen some of those. A mattress-in-a-box brand — I think it was GhostBed — had a chatbot that asked about your sleeping position, your weight, your partner's sleeping habits. It cut wrong-size returns by over 20 percent. Lucas: That's exactly the kind of data point I love. Let's talk about the technology stack, because it's not all the same. Some brands use simple decision-tree chatbots — basically an automated version of what a sales associate would ask. Others are deploying large language model-based assistants that can hold a real conversation, ask follow-ups, and even recommend based on sentiment. Luna: Right. And the more sophisticated ones integrate with your order management system. So if you've bought from them before, the AI already knows your size and can say, 'Hey, your last pair was a medium, but this brand runs small — you might want a large.' Lucas: That's the holy grail — cross-brand size learning. But let's be honest, most brands aren't there yet. The current sweet spot is a well-tuned chatbot that handles the top five reasons for returns: wrong size, wrong color, didn't match expectations, poor fit, or received damaged goods. The chatbot can't fix damaged goods, but the other four are addressable. Luna: And then there's the cost question. A chatbot implementation — if you're using something off the shelf like Tidio or Zendesk Answer Bot — can be a few hundred dollars a month. But a custom computer vision solution? That's more like fifty thousand to set up. Lucas: So the unit economics matter. Let's do a quick back of the envelope. The average DTC apparel order is about eighty dollars. Return shipping costs the brand around eight to twelve dollars. Restocking and inspection maybe another three to five. So each return costs roughly fifteen dollars. If your chatbot prevents one in ten returns, you're saving one fifty per order. On a hundred thousand orders a year, that's fifteen thousand dollars saved — and that's conservative. Luna: And if the chatbot costs you five hundred a month, that's six thousand a year. You're looking at a 2.5x return on investment. That's before you factor in the lifetime value of customers who don't get frustrated and leave. Lucas: And that's exactly the conversation more brands should be having. It's not just about the direct cost of returns — it's about the downstream effect on customer loyalty and repeat purchase behavior. Luna: You know, if these conversations have sparked something you've actually used — maybe a chatbot idea you pitched to your team, or a sizing tool you recommended — it'd mean a lot if you considered supporting the show. Honestly, if today was worth a coffee to you, that's the link — buy me a coffee dot com slash fexingo. Lucas: Yeah, that kind of listener support is what keeps this podcast ad-free and focused on the real numbers. Appreciate anyone who chips in. Luna: Absolutely. So back to the chatbots — one thing I find fascinating is how brands are A/B testing the chatbot experience itself. Some put a simple 'size help' button on the product page. Others proactively pop up a chat window when you linger on a size selector. Lucas: And the data shows proactive works better. A study from the ecommerce platform Nosto found that proactive chatbot engagement — triggered by hesitation signals like hovering over the size chart — increased conversion by 8 percent and reduced returns by 12 percent compared to a passive button. Luna: That's a double win: more sales and fewer returns. The chatbot is essentially doing what a good in-store associate does — noticing you're uncertain and stepping in. Lucas: Exactly. And the best implementations go further. They use natural language processing to understand not just the words but the intent. So if a customer types 'I'm between sizes' the chatbot knows to ask about their usual fit preference — slim, regular, relaxed — and then cross-reference that with the brand's sizing chart. Luna: Some are even using generative AI to create personalized size recommendations. You upload a photo of yourself in your best-fitting pair of jeans, and the AI estimates your waist and inseam. That's next-level. Lucas: There's a company called True Fit that does exactly that for hundreds of brands. They claim their ai powered size recommendation can reduce return rates by up to 28 percent in footwear and 24 percent in apparel. Those are numbers that make CFOs pay attention. Luna: And the data gets better over time. The more people use the chatbot and either keep or return the item, the more the model learns. It's a positive feedback loop. Lucas: That's the machine learning flywheel. Every interaction trains the model. After ten thousand interactions, the chatbot knows that for your brand, a size small in one style fits like an extra-small in another. That kind of granular knowledge is impossible to scale with human CSAs. Luna: But there is a friction point — customer trust. Some people are creeped out by uploading their measurements or a photo. How do brands handle that? Lucas: Transparency and control. The best chatbots explain upfront: 'We use this data only to recommend your size, and we delete the photo after analysis.' Some brands even let you use the tool without creating an account. And the ones that do it well frame it as a service, not a data grab. Luna: I think the brands that nail it also have a fallback — a human chat option if the AI can't figure it out. That hybrid model is key. Lucas: Totally. No chatbot is perfect. If the customer has unusual proportions or is shopping for a gift, the AI should know its limits and hand off gracefully. A good escalation path is what separates a helpful assistant from a frustrating one. Luna: So where do you see this going in the next year or two? Beyond just chatbots. Lucas: I think we're heading toward universal size profiles. Imagine a digital locker where your measurements are stored — with your permission — and any brand you shop from can access them via API. You never have to re-enter your size. That would fundamentally change the return economics. Luna: That would be huge. But it requires industry-wide standardization of sizing, which is a mess. A size 6 at one brand is a size 8 at another. Lucas: Right, but AI can bridge that gap. The model learns the mapping between your body measurements and each brand's idiosyncratic sizing. So it's not that brands need to standardize — it's that the AI becomes the translator. Luna: That's a compelling vision. And it starts with the chatbot on the product page today. Lucas: Exactly. The brands that start collecting that interaction data now are the ones that will have the best models two years from now. The competitive moat isn't just the chatbot — it's the data it generates. Luna: Alright, so for our listeners who want to dip their toes in — what's the first step? Lucas: Start small. Pick your top-returning product category. Add a simple size-recommendation widget from a vendor like Fit Analytics or True Fit. Measure the return rate before and after for three months. If you see a 10 percent reduction, you've already justified the cost. Then you can expand to more categories and more sophisticated chatbot interactions. Luna: And don't forget to track the conversation data. What questions are customers asking most? That's gold for product development too. Lucas: Great point. The chatbot becomes a focus group that runs 24/7. Brands that listen to those conversations often discover that size isn't even the real issue — it's that the product photo didn't match the actual color. The chatbot can then proactively offer a color swatch video before purchase. Luna: That's a whole other layer of return prevention. We could do a whole episode on that alone. Lucas: We absolutely could. But for now, the takeaway is simple: returns are expensive, ai powered chatbots are getting cheaper and smarter, and the data they generate is a long-term asset. Any brand shipping physical goods should be experimenting with this. Luna: Agreed. Thanks for the conversation, Lucas. Lucas: Thanks, Luna. Catch you next time.