Size and fit is the single largest driver of clothing, footwear, and activewear returns — consistently the most common return reason for fashion retailers and one of the most expensive, since apparel returns involve reverse logistics costs, processing time, and often discounted resale of returned items. AI fit recommendation tools attack this problem at its source: reducing the probability that a customer buys the wrong size by giving them better information at the point of purchase. Done well, they reduce return rates, improve purchase conversion (customers who are uncertain about fit often abandon rather than guess), and increase customer satisfaction.
This guide covers how the different categories of fit tools work, which tools are worth evaluating at different store scales, and how to implement them in a way that actually moves the metrics that matter.
Why Generic Size Charts Fail
Most ecommerce size charts present a simple table: Small fits chest 34–36″, waist 28–30″, and so on. This format fails customers in three systematic ways. First, customers often don’t know their own measurements in the units the chart uses — they know they’re “usually a medium” but not their chest measurement in inches. Second, size charts don’t account for fit preference — a customer who prefers a relaxed fit and one who prefers a slim fit may both wear the same measurement-based size but want different size recommendations. Third, sizing is inconsistent across brands, so “Medium” is not a transferable fact, making a generic size chart misleading for multi-brand retailers.
The consequence is that customers make size decisions based on incomplete information, return rates are elevated by fit-related purchases that could have been avoided with better guidance, and the customer’s experience of “I always have to guess and sometimes I get it wrong” becomes associated with the retailer rather than the inherent complexity of sizing.
👗 AI Fit Tools: What Each Approach Addresses
Tools by Scale and Sophistication
Kiwi Sizing is one of the most accessible fit recommendation apps for Shopify, offering dynamic size recommendations based on customer measurements and product size charts. It handles the common case — a customer wants to know what size to buy for a specific product based on their measurements — with minimal setup complexity. The measurement-based recommendation engine works across multiple product types and the interface is customisable to match the store’s design. It’s the appropriate starting point for stores moving beyond static size charts who want a meaningful improvement without significant implementation investment.
True Fit is a more sophisticated platform used by larger fashion retailers. It combines product measurement data with individual customer profiles — built from their size history, stated preferences, and purchase and return behaviour across the True Fit network — to produce personalised size recommendations that improve as the platform accumulates more data about each customer. The recommendations are more accurate than measurement-only tools because they incorporate actual purchase behaviour from customers with similar profiles. The implementation complexity and cost are proportionally higher, making it appropriate for retailers with meaningful transaction volume who have exhausted the gains available from simpler tools.
Fit Analytics (acquired by Snap) takes a similar approach to True Fit — using purchase and return data alongside product measurements to build predictive size recommendations. Its machine learning models improve continuously as more transaction data flows through the platform. Like True Fit, it’s most appropriate for retailers at scale where the accuracy gains from data-driven modelling justify the implementation investment.
Virtual Try-On: When It Helps and When It Doesn’t
Virtual try-on tools — which use computer vision and AR to show how a garment would look on the customer — address a different problem from size recommendation. Size recommendation answers “will it fit?” Virtual try-on attempts to answer “will it look the way I want it to?” These are both valid questions, but they’re different questions, and the evidence that virtual try-on reduces returns is more mixed than the evidence for size recommendation tools.
Virtual try-on is most useful for categories where aesthetics are the primary purchase driver — fashion-forward clothing, eyewear, accessories — and where the customer’s uncertainty is about how something will look rather than whether it will fit. It’s less useful for functional categories (athletic wear, workwear, basics) where fit accuracy matters more than aesthetic preview. The customer device requirement — quality front-facing camera, willingness to use an AR feature — also constrains the addressable audience to a subset of shoppers.
For most ecommerce retailers, size recommendation tools have a clearer and more direct return reduction pathway than virtual try-on. Starting with size recommendation and adding virtual try-on only for specific high-aesthetic categories is the more evidence-based prioritisation.
🚀 Implementing a Fit Recommendation Tool: Key Steps
The Data Feedback Loop
The fit tools that improve over time are the ones connected to return data. When a customer returns an item with the reason code “too small” or “too large,” and that return is associated with a recommendation the tool made, that data point tells the tool its recommendation was wrong for this customer profile and this product. Feeding that signal back into the model improves future recommendations for similar customers and similar products.
Most fit platforms handle this feedback loop automatically if connected to the store’s returns data — which requires a data integration between the fit tool and the returns management system. Setting up this integration from the start, rather than treating it as a future enhancement, means the model begins improving from the first return rather than waiting until the integration is built later.
Measuring Success Honestly
The primary metric for fit tool success is return rate in the target product categories, specifically the proportion of returns with fit-related reason codes. A tool that reduces fit returns by a meaningful proportion has paid for itself many times over in reduced reverse logistics costs, resale losses, and customer service overhead. Secondary metrics worth tracking: purchase conversion rate in categories where the fit tool is active (fit tools should increase conversion by reducing uncertainty), and the uptake rate of the fit recommendation widget (if customers aren’t using it, its impact is limited regardless of its accuracy). Review these metrics at the three-month and six-month marks after launch to assess whether the tool is producing its expected impact and to identify any aspects of the implementation that need adjustment.