On this page
Signing a sizing vendor is not a checkbox decision — it's a commitment that touches your product data, your checkout speed, and your return rate for years. Most merchants evaluate on price and demo polish, then discover six months in that the vendor can't produce an accuracy number, owns their fit data, or breaks under Black Friday load. This is the due-diligence list to run before you sign anything.
Why Sizing Vendor Evaluation Deserves a Formal Process
A size-recommendation tool sits in your critical purchase path. If it's slow, it hurts conversion. If it's wrong, it drives returns instead of preventing them. And if the contract is vague, you end up locked into a platform you can't leave without losing your historical fit data.
Treat this like any other RFP: written questions, written answers, and references you actually call. If you're earlier in the process, start with how to choose a size recommendation solution for the broader framework, then use this list for the vendor calls themselves.
The 12 Questions to Ask a Sizing Vendor
1. Who owns the fit and body-shape data we generate?
This is the single most consequential question. Some vendors treat every customer interaction as their proprietary training data, meaning you can't export it, can't take it to a competitor, and can't use it to train your own models later. Get contract language, not a verbal assurance.
2. Can you show accuracy numbers from a comparable merchant, not just a lab test?
Ask for a size-prediction accuracy rate and a return-rate reduction figure from a merchant in your category and price range. Lab benchmarks on synthetic bodies say little about how the model performs on your actual return-reason codes.
3. What happens to our data if we cancel?
Ask explicitly: is fit data deleted, exported, or retained by the vendor. This determines your real switching cost and should be spelled out in the offboarding clause, not left implicit.
4. How is customer body and measurement data secured and processed?
You need specifics — encryption in transit and at rest, data residency, sub-processor list, and whether biometric-adjacent data (height, weight, body shape) triggers extra regulatory obligations in your markets (GDPR, CCPA, and increasingly state-level biometric privacy laws).
5. What is the added latency on the product page?
A size widget that adds 300+ milliseconds to page load will show up in your Core Web Vitals and your conversion rate. Ask for real client latency numbers, not a marketing claim, and request a live product page to test yourself.
6. How does the recommendation engine handle new products with no size history?
Every vendor performs fine on your top sellers, which already have return data to learn from. The real test is a brand-new SKU with zero orders. Ask exactly what signal the model falls back to — garment measurements, category averages, or a generic guess.
7. Does it work across our multi-brand or multi-fit-model catalog?
If you sell brands from different factories with different fit blocks, a single "runs small / runs large" model will be wrong constantly. Ask how the vendor handles per-brand or per-style calibration.
8. What's the actual return-rate impact, and how is it measured?
Ask for the exact methodology: pre/post comparison, control group, or lift versus a holdout. Vague claims like "reduces returns" without a measurement method are marketing, not evidence.
9. What support and SLA do we get post-launch?
Get response-time commitments in writing, not "we're usually pretty responsive." Ask who owns fixes when the recommendation is visibly wrong on a live product page during a sale event.
10. How does pricing scale as we grow?
Understand whether pricing is per-session, per-order, or flat-tier, and get the number at 2x and 5x your current volume. Vendors that price attractively at your current size sometimes become unaffordable exactly when the tool is working and volume grows.
11. Can we run a pilot on a subset of the catalog before full rollout?
A vendor confident in their accuracy should have no problem running a 60-90 day pilot on one category with a clear before/after comparison. Refusal to pilot is itself a signal.
12. What does the integration and platform footprint actually look like?
Ask what changes on your storefront — theme edits, app blocks, or a full replatform — and how the integration behaves on your specific stack. If you're on Shopify, Ticimax, or ikas, get platform-specific answers, not a generic "we integrate everywhere."
Turning Answers Into a Comparison
Once you've run these questions with two or three vendors, score the answers side by side instead of relying on memory or gut feel.
| Question area | Weight | Vendor A | Vendor B |
|---|---|---|---|
| Data ownership & exit terms | High | ||
| Accuracy proof (comparable merchant) | High | ||
| Security & privacy specifics | High | ||
| Latency (measured) | Medium | ||
| New-SKU cold-start handling | Medium | ||
| Pricing at 2x/5x volume | High | ||
| Pilot availability | Medium |
Use this alongside the sizing vendor buyer checklist to make sure you're not missing implementation-stage questions this list doesn't cover, like design customization or A/B testing support.
Red Flags to Watch For During the Call
Vague answers to the data-ownership and accuracy questions are the two biggest tells. If a vendor can talk fluently about their UI but stumbles on "what happens to our data if we leave," treat that as a serious signal, not an oversight.
When to Walk Away From the Build Option Instead
If none of the vendors satisfy your data-ownership or accuracy-proof bar, it may be worth revisiting whether to build in-house. The comparison in build vs. buy for sizing tools lays out the real cost and timeline tradeoffs so that decision isn't made on price alone.
Why This Matters More Than the Demo
According to procurement research on SaaS buying, buyers who formalize due diligence around data rights and security terms before signing are significantly less likely to face a costly re-platforming project within two years — see McKinsey's research on SaaS procurement discipline for the broader pattern across enterprise software categories. Sizing tools are no exception: the cost of switching after a bad fit isn't just money, it's months of re-integration and lost fit-history data.
A polished demo tells you almost nothing about how a vendor performs at 2am during a flash sale, or what your options are eighteen months later when your catalog has tripled. The twelve questions above are designed to surface exactly the answers a demo won't give you.
Frequently Asked Questions
How long should a sizing vendor evaluation take?
Most merchants can complete a thorough evaluation, including reference calls and a pilot, in four to eight weeks. Rushing past the pilot stage is the most common mistake, since accuracy claims are hard to verify without live data on your own catalog.
Should we ask for references from merchants in a different category?
References from your own category and price point are far more useful than references from unrelated verticals, since fit behavior and return drivers differ significantly between categories like footwear, outerwear, and fitted apparel. If a vendor can't produce a same-category reference, ask why.
What if a vendor won't share their accuracy methodology?
Treat this as a disqualifying red flag rather than a minor gap. A vendor confident in its numbers will readily explain how they were measured; reluctance to do so usually means the numbers don't hold up under scrutiny or were generated in a lab setting far removed from real merchant catalogs.
Is data ownership really worth this much emphasis?
Yes — fit and body-shape data is some of the most valuable behavioral data your store generates, and losing access to it on cancellation means starting from zero with the next vendor. Merchants who skip this question often discover the cost only when they try to leave.