Size curve optimization: the Shopify fit playbook

Size curve optimization: the Shopify fit playbook

Size curve optimization means using AI to turn a customer photo and a garment’s real measurements into a personalised size recommendation and a photorealistic preview of the fit. For most Shopify merchants, the right first step is not a full rollout. It’s a two-week audit of your ten best-selling SKUs to check whether your product data can even support accurate sizing.

Before you touch any vendor, check three things in your own store:

• Product data quality — do you have real garment measurements (not just S/M/L labels) on file?
• Sample garment accuracy — has anyone actually measured the physical item, or are you trusting a supplier spec sheet?
• Instrumented pages — which product pages get enough traffic to produce a statistically useful pilot within a month?

Key Takeaways

Size curve optimisation cuts fit-related returns and lifts conversion only when garment-specific measurement data feeds an AI-driven virtual try-on placed at the actual purchase decision point.

Point Details Audit before you pilot Confirm real garment measurements and clean variant data exist before choosing a vendor. Place the widget correctly Position try-on beside add-to-cart, not in a buried tab, to maximise engagement. Measure return rate by reason Track fit-related returns separately from “changed my mind” to isolate real impact. Give pilots six weeks minimum Breakeven typically lands within six to fourteen weeks depending on traffic volume. GarmCheck runs the full stack Shopify app or JS snippet delivers eight-measurement sizing and a ten-second photorealistic render, with a 14-day trial to pilot on your own SKUs.

Table of Contents

• How does size curve optimisation actually work?
• What business results does size curve optimisation deliver?
• Shopify implementation: audit, pilot, scale
• Which KPIs prove size curve optimisation is working?
• Best practices and pitfalls that decide adoption
• What do early industry pilots actually show?
• What one implementer noticed on the store floor
• Why GarmCheck is built for this playbook
• Sources

How does size curve optimisation actually work?

Four components do the work. A body capture layer turns a single front-facing photo into a measurement set, typically covering around eight key points: chest, waist, hips, inseam, shoulder width, and similar. A garment-measurement layer holds the actual flat-measurements of each product, ideally by variant, not just by size label. A fit model compares the two, accounting for fabric stretch, cut, and intended ease, and outputs a recommendation. Finally, a photorealistic render shows the shopper what that recommendation looks like on their own body shape, inside the virtual try-on widget.

The data flow runs: photo uploaded → measurements extracted → matched against garment specs → size recommendation and visual render returned, usually in under ten seconds. On Shopify, this typically hooks into product metafields for garment dimensions, sits as a widget near the variant selector, and fires analytics events into your existing stack.

Most failures trace back to one thing: a fit model trained on generic body data rather than garment-specific mapping . Feed it a size chart without stretch or composition data and it will produce a confident, wrong render, which does more damage to trust than no try-on at all.

Pro Tip: Ask any vendor to show you a render on a garment with high stretch (like leggings) against one with none (like a structured blazer). If the fit recommendation logic doesn’t change between the two, the model isn’t garment-specific.

A simple three-step diagram (capture, match, recommend) is worth commissioning for internal buy-in decks. It’s the fastest way to explain the mechanism to a merchandising team that has never seen it before.

What business results does size curve optimisation deliver?

The commercial case rests on four levers: fewer returns, higher conversion, a lift in average order value, and stronger repeat purchase from shoppers who trust your sizing, as detailed in return-reduction strategies and profit impact . None of these move in isolation. A return that never happens also frees up the stock and staff time that reverse logistics would otherwise consume.

The evidence so far: Zalando’s own trial recorded a 40% drop in returns during an April 2023 test. DressX intelligence found try-on users were roughly 50% more likely to purchase than non-users, with sharper lifts among luxury shoppers. Separate vendor case data points to conversion improvements in the low-to-mid teens for merchants adopting photorealistic fit tools.

Take a mid-market store selling 2,000 units a month at an average return cost of £12 per item (shipping, restocking, damaged-stock write-offs).

Fit-critical categories benefit most:

• Dresses, jeans, and jackets, where cut and ease vary wildly between brands
• Activewear, where stretch and compression fit drives satisfaction
• Basic tees and accessories sit lower priority, since sizing variance is smaller

Shopify implementation: audit, pilot, scale

Run this in four phases rather than one big-bang launch.

Phase 1: Audit (1–2 weeks). Check product-data completeness (do you have real measurements per variant, and fabric composition?), variant hygiene (are sizes labelled consistently across the catalogue?), return-reason tagging (can you isolate “too tight” from “changed my mind”?), and whether your analytics can capture a new event type.

Phase 2: Pilot (4–6 weeks). Choose 10–25 SKUs across your fit-critical categories. Define your KPIs upfront. Design a proper A/B split rather than a soft launch to everyone. Make sure sample garments used for measurement data are physically checked, not pulled from supplier PDFs.

• Install the Shopify app or JavaScript snippet
• Map garment measurements to product metafields
• Place the try-on widget near the add-to-cart button, not buried in a tab
• Connect the CRM (Klaviyo or similar) for post-purchase and abandoned-fit follow-up
• Confirm analytics events are firing correctly before scaling traffic

Phase 3: Iterate (2–4 weeks). Review early data, fix any garment-mapping errors, and widen the SKU set.

Phase 4: Scale. Roll out store-wide once the pilot categories show consistent uplift. Operator data suggests breakeven typically lands somewhere between six and fourteen weeks depending on traffic volume, so give your pilot at least six weeks before drawing conclusions.

Pro Tip: Document photo-consent language in your privacy policy and directly on the try-on widget itself, not just buried in your terms page. Shoppers upload a photo far more readily when they see, in plain language, that it isn’t stored or shared.

Which KPIs prove size curve optimisation is working?

Track five numbers, in this order of priority: product-page conversion rate, try-on engagement rate (percentage of visitors who complete a try-on), add-to-cart rate for try-on users versus non-users, product return rate segmented by reason, and average order value for try-on users. Use a rolling 30-day cohort for early signal, then confirm with a 90-day window before making permanent decisions, since seasonal buying patterns can distort a shorter cut.

Metric Pre-pilot baseline Expected direction during pilot Product-page conversion rate Current store average Increase Try-on engagement rate N/A (new metric) Establish baseline, then grow Fit-related return rate Current return-reason data Decrease Average order value Current store average Slight increase

Run try-on versus no-try-on as your control split, not old-traffic versus new-traffic, or seasonality will contaminate the result.

Best practices and pitfalls that decide adoption

Get the fundamentals right and adoption follows. Get them wrong and even good technology looks like it failed.

What works: authoritative garment specs measured from physical samples, a try-on widget placed directly beside the add-to-cart button, explicit consent language, coverage across the full size range rather than just sample sizes, and a render that loads in seconds, not minutes.

What derails pilots:

• Relying on a generic body model instead of garment-specific fit data
• Product metadata gaps that force the fit model to guess
• Burying the widget in a tab shoppers never open
• Customer service teams left untrained on how to answer fit questions

Pro Tip: Give your CX team a one-page cheat sheet on how the try-on recommendation logic works. The single fastest way to lose shopper trust is a support agent who can’t explain why the tool suggested a size.

Any of these means your garment data, not the shopper, is the problem.

What do early industry pilots actually show?

Breuninger’s six-week “be your own model” test, run over Black Week and the holiday season, showed shoppers using personalised selfie-based try-on converted at a higher rate and produced stronger contribution margin than the control group. Zalando’s applied science team built what it calls a “ fit intelligence ” layer, feeding body-to-garment mismatches back into design decisions rather than treating fit as a one-off fix.

Expect variance by category and by pilot maturity. Early tests often show the largest swings, dresses and jackets especially, because that’s where fit failure was worst to begin with. Numbers settle as the recommendation model matures and the SKU set widens.

What one implementer noticed on the store floor

The biggest lesson from watching pilots run isn’t technical. It’s placement. Move the try-on widget from a product-description tab to right beside the “Add to Cart” button and engagement jumps immediately, because that’s the exact moment a shopper is deciding whether to trust the size.

Why GarmCheck is built for this playbook

GarmCheck runs the whole size curve optimisation stack described above inside a Shopify app or JavaScript snippet, so you’re not briefing engineers to build a fit model from scratch. Upload a single front-facing photo and shoppers get a photorealistic render in under ten seconds, built from eight body measurements matched against your actual garment data.

Klaviyo integration means try-on data becomes a CRM asset, not a one-off interaction, feeding follow-up campaigns to shoppers who tried an item but didn’t buy. GarmCheck handles the measurement layer, the garment mapping, and the analytics events, so your pilot phase focuses on picking the right SKUs, not building infrastructure. A 14-day free trial lets you run the exact pilot outlined above on your own top sellers before committing. For internal sign-off, GarmCheck’s one-pager is built to hand straight to a stakeholder who needs the business case in five minutes. Start with a look at the virtual try-on product itself and map it against your audit checklist this week.

Sources

• How Breuninger boosted sales with its “be your own model” AI | Google Cloud Blog
• How Zalando uses technology to help customers find the right size | Zalando SE
• Retailers rely on virtual try-on to curb returns and boost conversions | eMarketer
• AI virtual try-on and luxury: DressX intelligence | Business of Fashion

Recommended

• Plus size try-on for Shopify merchants: a practical guide — GarmCheck
• Denim fit guide for UK Shopify merchants — GarmCheck
• Virtual Try-On for Fashion Retailers | GarmCheck
• Body measurements from photo: a practical guide for UK fashion retailers — GarmCheck