YundaDeskSize questions look like small pre-sale chats until the return bill arrives. A customer orders one size too small, asks for an exchange, then needs shipping instructions, inventory checks, and follow-up. For apparel and footwear brands selling across markets, the same letter size can mean very different things to customers in the US, Europe, Japan, Southeast Asia, or the Middle East.
That is why a strong size recommendation script should never stop at “choose your usual size.” A better workflow combines support prompts, size charts, fit notes, fabric behavior, and clear escalation rules. AI can answer the repetitive questions first, while human agents step in for unusual body types, complex foot shapes, or return disputes.
Build a Size Knowledge Base Before Writing Scripts
Many stores already have a size chart, yet agents still struggle to answer fit questions. The problem is that a chart lists measurements; it does not explain how customers should use them. For every key SKU, prepare these details in the knowledge base:
| Detail | Knowledge base entry | Customer question it supports |
|---|---|---|
| Core size chart | S/M/L mapped to bust, waist, garment length, shoulder width, or insole length | “I am 168cm and 60kg. What size should I choose?” |
| Fit profile | Slim, regular, relaxed, cropped, narrow toe box, low instep | “Will this feel tight?” |
| Fabric behavior | No stretch, slight stretch, high stretch, shrinkage after wash | “What if I am between sizes?” |
| Model reference | Height, weight, size worn, fit effect | “The model wears M. What about me?” |
| Return policy | Whether size exchanges are allowed and who pays shipping | “Can I exchange it if the size is wrong?” |
YundaDesk can build a knowledge base from uploaded documents, crawled website pages, and manually written Q&A. The important part is not only storing the table. You also need to include the way customers actually ask: “I wear Zara M,” “my feet are wide,” “I want an oversized look,” or “I am between two sizes.”
Size and Fit Recommendation Scripts for Apparel and: tier incoming questions by judgment risk
Ask for Height, Weight, and Fit Preference First
When a customer asks “What size should I get?”, do not jump straight to a recommendation. First collect the variables that affect the answer most: height, weight, usual size, size system, and preferred fit.
Use a prompt like this:
I can help you choose a safer size. Could you share your height, weight, usual size, and whether you prefer a fitted or relaxed look?
For pants, please also share waist or hip measurement. For shoes, please include your usual EU, US, or UK size.
Once the customer replies, recommend with context:
Based on 165cm / 55kg and your usual M, I would start with M for this item. It has a regular fit, so the shoulder and bust should not feel too tight.
If you prefer a looser look or plan to layer underneath, choose L. If you want a closer fit, M is the better option.
This works better than “choose M” because it explains the reasoning. When AI pulls from the knowledge base, the response should follow the same structure: confirm the input, give the primary recommendation, then explain when to size up or down.
Use Usual Size Carefully Across Brands
Cross-border customers often ask with a reference brand or size: “I usually wear M,” “I wear Nike 8,” or “I wear Zara S.” That is useful context, but it is not a direct conversion. Brand fit, country sizing, and product category all vary.
Use this script:
Your usual size is a helpful reference, but different brands fit differently. To avoid choosing the wrong size, I recommend checking this item’s actual measurements.
If you usually wear M and your bust is around 88-92cm, with shoulder width close to our M chart, M should be the safer choice. If M often feels tight at the shoulders, consider L.
For footwear:
If you usually wear US 8, EU 39 is a good starting point. This style has a slightly narrow toe box.
If your feet are wide or your instep is high, consider going up half a size. If your feet are slim, your usual size should work.
The key is to avoid absolute promises. Say “safer,” “recommended,” or “if you prefer X, choose Y.” Do not say “this will definitely fit.” A size recommendation script reduces uncertainty; it does not remove it completely.
Turn the Size Chart Into Plain Instructions
Many returns happen not because the chart is wrong, but because customers do not know how to measure. Your support script should explain how to measure, especially for pants, dresses, lingerie, and shoes.
| Category | Script |
|---|---|
| Tops | “Measure a top that fits you well. Lay it flat and compare bust, shoulder width, and length with our chart.” |
| Pants | “Check both waist and hips. If the waist fits but your hips are near the upper limit, choose the larger size.” |
| Dresses | “Check bust, waist, and length together. If one area is much tighter than the others, choose based on that area.” |
| Shoes | “Measure foot length, add 0.5-1cm for movement space, then compare it with insole length.” |
YundaDesk’s AI Agent is well suited for this standardized guidance. It can ask for missing details first, pull the matching product’s size knowledge, then give a recommendation with reasoning. Agents no longer have to paste the same measuring instructions all day.
Handle Between-Size Cases With Rules
“I am between M and L” is one of the most common fit questions in apparel support. Do not decide only by weight. Use fit type, fabric, and preference:
- Slim fit: when between sizes, the larger size is usually safer for movement.
- Relaxed fit: if the customer does not want an oversized look, the smaller or usual size may work.
- No-stretch fabric: size up, especially when shoulder, bust, or hip measurements are near the limit.
- High-stretch fabric: the smaller size can work for a closer fit, but explain the effect.
- Narrow toe box: wide feet, high instep, or thick socks usually mean going up half a size.
Use this script:
You are between M and L. Since this item has limited stretch and a slimmer fit, I recommend L for more comfortable movement.
If you prefer a very fitted look and your bust or shoulder measurement is not near the upper limit of M, M can also work.
These answers should become standard scripts. If agents later find that a specific SKU gets fewer exchanges when customers size up, they can correct AI and turn that experience into a learning suggestion. In YundaDesk, “gets smarter over time” means the update is confirmed by the merchant first, then becomes active. Every learning step is traceable, testable, and revertible.
Ask Footwear Customers About Length, Width, and Use Case
Shoe size recommendations create more disputes than many apparel questions because fit depends on foot length, width, instep height, sock thickness, and use case. Ask in this order:
- Usual size system: EU, US, UK, or JP.
- Measured foot length in centimeters.
- Foot shape: wide, slim, or high instep.
- Use case: sports, commuting, long walks, or thick socks.
Then answer like this:
Based on your 24.5cm foot length, I would start with EU 39 for this style.
The toe box is slightly narrow, so if your feet are wide or you wear thick socks in winter, EU 40 is safer. If your feet are slim, EU 39 should feel more fitted.
For kids’ shoes, athletic shoes, and boots, add extra notes about growth room, sock thickness, and shaft circumference. Do not force every shoe into one template. The knowledge base should keep separate scripts by footwear category.
Let AI Recommend First and Humans Back Up
Size recommendation is a strong fit for AI-first support because the questions repeat, the answer depends on product knowledge, and customers expect fast replies. But it should not be fully unsupervised. Define the boundaries clearly:
| Scenario | Handling rule |
|---|---|
| Complete information and matching size chart | AI recommends directly and explains the basis |
| Missing information | AI asks for height, weight, usual size, or preference first |
| Unusual body type or complex foot shape | AI gives an initial suggestion and offers handoff to human |
| Customer asks for a guarantee | AI explains the basis and return policy without making an absolute promise |
| Refund, compensation, or price change | Must hand off to human approval |
Whether the customer asks through a website widget, WhatsApp, Instagram, TikTok, or email, YundaDesk brings the conversation into one workspace with the same customer profile and the same knowledge base. If the customer switches channels, agents still see the context instead of asking them to repeat everything.
For more on where AI should stop and agents should step in, read AI answers first, humans back up. For apparel and footwear teams, clearer boundaries make agents more confident about letting AI handle high-volume fit questions.
A useful size recommendation script does not push every customer into a single size. It explains the decision process: what to ask, what to compare, why one size is recommended, and when another size is safer. Put that logic into the knowledge base, and AI can handle most repetitive size questions while human agents focus on edge cases, return disputes, and high-risk decisions.