Guides
Prepare your Shopify catalog before you launch the assistant
Catalog quality is the launch gate. Incomplete variants and missing attributes produce poor guidance, not “smarter AI.”
A shopping assistant can only describe, recommend, and compare what the connected catalog actually contains. If variants are incomplete, titles are inconsistent, or availability is stale, the experience will feel unintelligent even when the product is working as designed. Catalog preparation is therefore a launch gate, not a polish step you schedule after the widget is live.
This guide is for merchandising, ecommerce operations, and the person who will own Shopify admin during a WardrobeIt launch. It does not replace the Help Center connection runbook. It explains what “ready” means in fashion and visual commerce terms, and how to inspect the catalog before shoppers ever see the assistant.
Why fashion catalogs fail conversational guidance
Apparel decisions combine product type, size, color, fabric, occasion, and price. Those ideas only become usable if they exist as data: variant options, product type, tags, metafields, collections, and images. A PDP paragraph that says “perfect for summer weddings” is helpful to a human and almost useless to a system if nothing in the record encodes occasion, length, or formality.
Common failure modes before launch include missing size curves on otherwise complete styles, color names that do not match swatch images, collections that mix gift cards with garments, and inventory that is not synced for the locations you actually sell from. None of those problems is “the AI being dumb.” They are catalog problems that conversation will amplify.
What “ready” means operationally
Ready does not mean every historical SKU is perfect. Ready means a merchandiser and an operator can both explain, for the products shoppers will actually meet, what the assistant is allowed to say. If you cannot defend a claim from a field, an image, or an approved knowledge entry, the assistant should not say it.
Treat unpublished, draft, and placeholder products as out of scope. A connected catalog that includes test titles, $0 prices, or “TBD fabric” copy will leak those facts into conversation. Clean the published assortment you intend to sell through, then connect.
Minimum record quality
- Identity: a unique, shopper-readable title; vendor; product type that matches how you merchandize, not an internal dumping ground.
- Sellable options: size, color, and any other shopper-facing option you sell. A single “Default Title” variant is a warning in fashion.
- Media: at least one live image per style, and variant images that match the option the shopper will select.
- Availability: inventory tracked where you intend the assistant to respect stock.
- Price: values you are willing to show in guidance. Do not leave placeholder prices on published SKUs.
What to inspect before you connect
Do this inspection in Shopify admin and on the live theme, not only in a spreadsheet export. Conversation will inherit whatever the shopper can also verify on the PDP. If admin and storefront disagree, fix that disagreement first.
- Product completeness: title, description, vendor, product type, and at least one live image per style.
- Variants: size, color, and any other shopper-facing option. Confirm option names are consistent (Size vs Sizes, Colour vs Color) across the assortment you will expose.
- Availability: inventory tracked for the locations and channels you sell from. Selling an unavailable size in conversation trains shoppers to distrust later answers.
- Collections and types: the buckets you will use for discovery, complete-the-look, and merchandising rules. Mixed “everything” collections make pairing harder to govern.
- Price and compare-at: values you are willing to show. Seasonal markdowns should match what finance has approved to display.
- Status and sales channels: unpublished or channel-hidden products should not be treated as fair game for storefront guidance.
A practical walkthrough on five SKUs
Do not audit the entire catalog on day one. Pick five SKUs that represent how you actually sell: a hero dress, a size-sensitive bottom, a color-heavy knit, an accessory, and a new-season style. For each SKU, open the Shopify product, the live PDP, and write down every question a shopper might ask. Then mark which answers exist in structured data versus only in a lifestyle caption.
Typical questions: What sizes exist? Which colors are in stock? What is the fabric? Is this midi or maxi? Can it be worn to a daytime event? What ships with it? If three of the five SKUs cannot answer “what sizes exist” or “is this in stock” from catalog fields, pause the widget install. Fix the records. Re-test.
How to score each SKU
- Pass: a merchandiser can point at the field or image that supports the answer, and the live PDP agrees.
- Gap: the answer exists only in a lookbook caption, a founder’s memory, or a campaign brief.
- Conflict: admin, PDP, and a policy page disagree. Conflicts are launch blockers, not footnotes.
The Help Center steps for connecting Shopify assume you are pointing WardrobeIt at a catalog you trust. Sync copies the current truth; it does not invent a better one. See Connect a Shopify Store only after this five-SKU pass is honest.
Images and attributes travel together
Visual commerce depends on images that match variants. If the black dress PDP shows a navy photo, conversational guidance will inherit that confusion. Align featured images and variant images before you ask the assistant to “show me the black one.” Color names should be stable enough that a shopper and a merchandiser mean the same swatch.
Lifestyle images are useful for humans and dangerous as the only source of pairing truth. If a lookbook photo includes a belt you do not stock, do not treat the photo as a complete-the-look specification. Pairing belongs to SKUs, not to borrowed styling.
Collections, types, and the discovery surface
Conversation will retrieve against the structure you already maintain. If dresses live under three product types and five overlapping collections, guided journeys will feel noisy. You do not need a perfect taxonomy to launch, but you do need a merchandiser to name which collections are allowed in guided shopping and which are operational dumping grounds (clearance mixed with gift cards, for example).
Write that decision down. It becomes the input for recommendation rules later. Catalog-only recommendations cannot compensate for a collection strategy that was never intended for shopper-facing retrieval.
Inventory as a conversation constraint
Fashion inventory moves. If an assistant recommends an out-of-stock size, the store trains shoppers to distrust every later suggestion. Decide whether you sell from one location or many, whether you allow backorder language at all, and whether location-level stock matches the channel the shopper is on.
If backorders exist, they belong in merchant knowledge and must match operations—not in an improvised assistant sentence. If they do not exist, the assistant should not invent a restock date.
Failure modes to catch before go-live
- Hero styles with missing size options on some colors.
- Color option values that do not match variant image filenames or alt text in any useful way.
- Published products with “lorem” descriptions or internal vendor codes as titles.
- Collections still containing last season’s SKUs that merchandising considers retired.
- Compare-at prices that imply a promotion finance has not approved to display.
- Gift cards, wrapping, or non-apparel add-ons mixed into apparel collections used for pairing.
Verification after you connect
After sync, repeat the five-SKU test inside the assistant before you spend on paid traffic. Ask for the size curve. Ask for the in-stock color. Ask for a complementary item only if that item is a SKU. Compare every answer to Shopify and the live PDP. If they disagree, stop debating the model and fix sync or source data.
Follow Getting Started for the launch sequence around this work. Catalog connection is one step; it is not the whole launch.
When the catalog is ready
You are ready to connect when those five SKUs can be explained without folklore. Then complete Connect a Shopify Store. After sync, keep a weekly habit: pick three in-demand styles and verify recommended variants against Shopify inventory. Seasonal drops deserve the same ritual, not a one-time launch audit.
Related reading: Shopify Catalog Sync, Catalog-Only Recommendations, Fashion, and the Guides hub for sibling operator articles.
