On this page
Most merchants install a size recommender, watch traffic to the product page tick up slightly, and then have no real answer when finance asks "did this actually reduce returns?" Without a defined KPI dashboard, a size recommendation tool becomes a widget nobody can defend in a budget review. This post lays out the four metrics that matter most, how to instrument each one cleanly, and how they connect to the return-rate story leadership actually cares about.
Why Size Recommendation KPIs Get Ignored
Fit tools are usually owned by whoever installed the app — often a UX or CX team member — while the return-rate P&L sits with finance or operations. That organizational gap means the tool's usage data and the return data never get joined in one report. The result: teams optimize for "is the widget rendering" rather than "is the widget changing purchase behavior in a measurable way."
The fix isn't more data. It's fewer, better-defined metrics tracked consistently. If you're building this out with your team, pair this framework with a proper fit and sizing reporting setup so the numbers live somewhere durable, not just in an app's internal dashboard.
The Four Core Size Recommendation KPIs
1. Adoption Rate
Adoption rate is the percentage of product-page sessions where a shopper actually interacts with the size recommender — clicks it open, enters a measurement, or completes a body-profile flow. This is your top-of-funnel health check. Low adoption usually means placement or visibility problems, not a bad recommendation engine.
Benchmark ranges vary by category, but a widget placed prominently near the size selector typically sees adoption in the 15-35% range of product-page visitors who reach the size selection step. If you're under 10%, investigate button copy, placement above the fold, and whether the CTA competes visually with "Add to Cart."
2. Follow Rate (Recommendation Acceptance)
Follow rate measures how often a shopper who receives a size recommendation actually selects that size at checkout, rather than overriding it. This is the single most important signal that your algorithm is trusted. A high adoption rate paired with a low follow rate tells you shoppers are engaging with the tool but not believing its output — often a sign the recommendation logic needs recalibration against your actual size charts.
Instrument this by tagging the recommended size at the moment it's shown, then comparing it against the size added to cart in the same session. Anything under 60% follow rate is worth a size-chart audit.
3. Size-Related Return Delta
This is the metric that ultimately justifies the tool's cost. Compare the size-related return rate (returns tagged "too small," "too big," or "doesn't fit") for orders where the recommender was used versus orders where it wasn't, over the same time window and product set.
| Metric | Non-recommender orders | Recommender-assisted orders |
|---|---|---|
| Size-related return rate | Baseline (e.g. 22%) | Typically 8-15 points lower |
| Average resolution cost per return | Full reverse-logistics cost | Same, but fewer instances |
| Repeat purchase within 90 days | Lower | Higher |
To calculate this cleanly, you need order-level tagging that persists from the recommendation event through to the return reason code — most WMS or OMS platforms support custom order tags for exactly this purpose. For the finance-facing version of this comparison, see how to prove size recommendation ROI to your finance team.
4. Average Order Value (AOV) Lift
Shoppers who trust a size recommendation are more willing to buy multiple items, including items from categories they'd normally avoid (fitted styles, cross-category bundles). Track AOV for recommender-assisted sessions against control sessions, controlling for promo exposure and traffic source to avoid conflating the two effects.
Instrumenting the Dashboard Correctly
Event Tracking Requirements
At minimum, your analytics layer needs to capture four events per session: widget impression, recommendation shown (with the recommended size value), size selected at add-to-cart, and order completion with the final size. These four events, joined on session ID, let you compute adoption, follow rate, and return delta without any manual reconciliation.
Avoiding Common Measurement Errors
The most common mistake is measuring return rate against all orders rather than a matched cohort — comparing recommender users (who are often more engaged, higher-intent shoppers to begin with) to all other traffic overinflates the tool's apparent impact. Use an A/B holdout or, at minimum, control for traffic source and product category when calculating the return delta. It's also worth reviewing this alongside broader try-on success KPIs if your fit tool includes a virtual try-on component, since try-on engagement and size adoption often move together but measure different behaviors.
According to industry analysis, apparel returns remain one of ecommerce's largest hidden costs, with sizing and fit issues cited as the leading driver — a pattern documented in retail analytics coverage from McKinsey on the operational cost of returns. That's the baseline this KPI dashboard is designed to move.
Turning KPIs Into Action
A dashboard only earns its keep if it changes decisions. Set a monthly cadence: review adoption and follow rate weekly (these move fast and reflect UX/placement changes), and review return delta and AOV lift monthly or quarterly, since return data lags order data by the length of your return window.
When follow rate drops for a specific category, that's your cue to check size-chart accuracy for that category specifically rather than assuming the whole engine needs retuning. When adoption is strong but return delta is flat, the recommendation logic itself — not the widget — is the problem to investigate.
Frequently Asked Questions
What's a good follow rate for a size recommendation tool?
A follow rate above 70% generally indicates shoppers trust the recommendation and the size charts underlying it are accurate. Rates between 50-70% suggest partial trust and are worth investigating by category. Anything below 50% usually points to a size-chart calibration problem rather than a UX issue.
How long should I wait before measuring return delta?
You need at least one full return window — typically 30 to 90 days depending on your policy — plus enough order volume per cohort to be statistically meaningful. Measuring too early will understate the return delta because returns haven't had time to post.
Can I measure these KPIs without a dedicated analytics tool?
Yes, as long as your ecommerce platform supports custom order tagging and you can join session-level events to order-level return reasons. Many merchants start with a simple spreadsheet pulling from Shopify or Ticimax order exports before graduating to a dedicated reporting layer.
Does AOV lift apply to every product category?
No — AOV lift tends to be strongest in fitted or size-sensitive categories like denim, activewear, and outerwear, where uncertainty about fit is the biggest barrier to adding a second item to cart. Commodity or one-size categories typically show minimal AOV movement from a size recommender.