YundaDesk“I usually wear a 9, what size should I order” - six out of ten messages in a footwear queue are asking exactly this, just phrased six different ways. For shoes and bags, the bulk of pre- and post-sale questions aren’t about product quality. They’re about fit, whether the color matches the photo, and whether the material will hold up. That repetition is exactly what AI should be handling first.
The hard part isn’t volume, it’s judgment. The same size label means different things across lasts, seasonal cuts, and leather batches. One careless “true to size” reply and your return rate spikes. This playbook breaks fit guidance, material care scripts, and return steps into templates you can hand to AI and agents, so “doesn’t fit” turns into a repeat purchase instead of a bad review.
Sizing questions: ask about foot shape before giving advice
The easiest way to get sizing wrong is treating a generic conversion chart as a universal answer. The same EU size behaves differently depending on whether the last is narrow, standard, or wide, whether the toe box is round or pointed, and whether there’s elastic in the design.
Structure each shoe or bag’s knowledge entry like this, instead of one blanket size chart:
| Field | What to include |
|---|---|
| Last description | Narrow / standard / wide, suited foot shape |
| Size conversion | Local size to EU/US, down to half sizes |
| Common follow-ups | “Will this fit a high instep?” “What socks work with this?” |
| Suggested response | Usual size + guidance based on this last’s fit |
| Escalation trigger | Unusual foot shape (bunions, wide feet, high arch), explicit request for a human |
With this, AI can give a real recommendation - your usual size plus how this particular last runs - instead of a flat “size up.” If the customer keeps pressing on details or has an unusual foot shape, escalate to an agent. For how to structure a knowledge base AI can actually draw on, see the knowledge base that feeds AI.
Material care: say it upfront, cut post-sale volume in half
Another major source of footwear and bag tickets is care instructions: does leather need conditioning, can canvas be machine washed, will metal hardware tarnish. Customers usually ask after they’ve already bought, and “it’s in the manual” only makes things worse at that point.
Build entries by material, covering three layers:
- Routine care: how often to clean, what tools to use or avoid (alcohol, direct sun, aggressive brushing)
- Common issues: whether seam separation, fading, or hardware oxidation counts as a defect or normal wear
- Warranty boundaries: what counts as a workmanship issue versus wear and tear that’s out of scope
| Material | Common question | Suggested AI direction |
|---|---|---|
| Full-grain leather | Will it crack, does it need conditioning | Recommend periodic conditioning in dry climates, avoid direct sun and rain, include a care frequency guideline |
| Canvas / nylon | Can it go in the washing machine | Recommend hand-washing and air-drying, machine washing may affect coating and shape |
| Plated hardware | Will it discolor or tarnish | Avoid prolonged contact with perfume or sweat, mild oxidation is typically normal wear |
Proactive care reminders like these are also a good fit for Yuna to help teach the AI what agents already know. When an agent adds a more accurate care tip mid-conversation and corrects the AI, that generates a pending learning suggestion - it only takes effect once you approve it, and it’s reversible. See how YundaDesk gets smarter over time for the full loop.
Proactive reminders: AI checks in before care issues turn into complaints
One of the biggest repeat-purchase opportunities for footwear and bags is reaching out before a care or condition issue turns into a complaint - a reminder to condition full-grain leather shoes a month after delivery, or a heads-up about moisture protection for canvas bags before the rainy season. Done well, this heads off complaints early. Done poorly, it reads as spam.
YundaDesk’s proactive outreach runs behind six guardrails that can’t be switched off, which is what makes it safe to let AI initiate:
- Cooldown period: no repeated outreach to the same customer in a short window
- Frequency cap: outreach per customer is capped over time
- Quiet hours: nothing sent during the customer’s local late-night hours
- No interrupting active chats: outreach pauses while a customer is mid-conversation with an agent
- Do-not-disturb list: customers who’ve asked not to be contacted stay off the list
- Sensitive actions always route to a human: anything touching refunds or compensation never sends automatically
Care reminders for footwear and bags are relatively low risk, so it’s reasonable to start in observe-only mode, confirm the timing and tone work, then move up to “confirm each message” or full automation. For the complete logic on guardrails and modes, see proactive outreach without annoying customers.
Return and exchange steps: spell it out so nobody has to ask twice
A large share of footwear and bag return questions is really about process: where to send the return, whether tags need to stay attached, whether the shoebox counts as an accessory, what the exchange flow looks like. When the steps aren’t clear, customers ask again and agents repeat themselves.
Break the return flow into a checklist AI can reference directly:
| Scenario | Suggested response | Escalation trigger |
|---|---|---|
| Wants to exchange for a different size | Provide the order number and target size; if in stock and the item is unworn with box and tags intact, we can start the exchange | Target size out of stock, past the exchange window, price difference involved |
| Shoes don’t fit after wearing once | If the item is unworn outdoors with box and accessories intact, and within the return window, a return can be submitted; confirm whether it’s been worn outside | Visible wear, past deadline, customer is upset |
| Hardware on a bag discolored - is that a defect | Please share the purchase date and photos; we’ll assess based on usage time and care to determine if it’s a quality issue | Needs a liability call, involves compensation, customer wants it in writing |
Clear return steps leave agents with the judgment calls
Cross-border context: default sizing systems differ by country
The easiest mistake in international footwear and bag support is copying the same sizing script across every market. European customers think in EU sizes, US customers in US sizes, and some Southeast Asian markets mix in centimeter measurements - a small conversion error is enough to trigger a return.
Country, language, and time zone are built into the cross-border CRM by default, no extra setup required. That means AI can phrase sizing advice in whatever system matches the customer’s country, instead of making them convert it themselves or asking “do you mean EU or US” every time. Whether the same customer messages through the website, WhatsApp, or Instagram, the system recognizes one unified customer profile, so sizing history doesn’t need to be re-explained just because they switched channels.
The same applies to language: AI automatically responds in whatever language the customer is using, so you don’t need to prepare a separate script per language - as long as the underlying sizing and care information in the knowledge base is accurate.
Test before launch: run it against real questions
A sizing and care knowledge base isn’t launch-ready just because it’s written. Test it against real historical questions, not just clean sample phrasing.
- Size conversion tables organized per style, not one generic chart
- Pull at least 10 real historical sizing questions for testing
- Include casual, colloquial phrasing (like “will my big toe hit the front?”)
- Confirm unusual foot shape or body type questions escalate correctly
- Check that AI isn’t inventing unconfirmed warranty promises
- Verify return steps match the actual policy word for word
Focus on three things: whether the sizing advice is grounded in real data, whether the care scripts overstate warranty coverage, and whether the return steps genuinely reflect current policy.
Footwear and bag support doesn’t improve by adding more agents - it improves by structuring the three high-frequency questions (how to size, how to care, how to return) so AI can handle the repetitive volume while agents focus on unusual foot shapes and liability calls. Get this workflow right, and “doesn’t fit” stops being a one-time bad review and becomes the setup for a more accurate next purchase.