Turn vendor information into usable interior design product records

The hard part is knowing what each field means

Designers work across catalogs, product pages, order confirmations and invoices. All may describe the same item differently. Before AI can reduce the copying, the studio needs a reliable way to identify the product and interpret the source.

Our Marais Home work includes vendor-specific product intake for a retail catalog: descriptions, images, pricing, SKUs, inventory and shipping rules. The reusable lesson is that one universal instruction is rarely enough. A vendor’s order confirmation and invoice can use different meanings for a similarly labeled price field.

For a design studio, the destination might be a selection library or project schedule rather than a store. The approach below is an adaptation: gather facts into a consistent record, preserve the evidence and review exceptions before those records affect quotes or orders.

Build a record that keeps facts separate

A useful starting record contains vendor, vendor SKU, product name, variant, dimensions with units, finish, source link, source date and image reference. Store price type and currency beside any price. Keep trade cost, published retail and the studio’s client price distinct; none should be inferred from the others without an approved rule.

Record quantity and unit of sale separately. One carton, one set and one individual item are different things. Inventory should describe what is actually available in the system you are maintaining, not the quantity shown on an incoming purchase order.

Keep a short vendor rule sheet: which document supplies the SKU, how variants are named, which price field is authoritative and who resolves discrepancies. At Marais, documenting these distinctions became part of the work itself. An exception discovered once should become a reusable check.

A worked example: the chair that almost matches

Fictional example: a selection note says Alder dining chair, natural. The vendor confirmation lists ALD-NAT-02, while the product page initially displays ALD-WAL-02. The photos look similar, but the second SKU is the walnut finish. A title match alone would attach the wrong image and specification.

The intake should keep the confirmed SKU, mark the finish and image as needing verification, and show both sources. A reviewer selects the matching variant before the record moves into the approved library. AI can gather candidate information, but confidence in a visual resemblance is not evidence that two variants are identical.

Suppose the same confirmation orders three sets of two chairs. Preserve that unit of sale and explicitly convert it to six individual chairs only where the destination requires individual units. Otherwise, an apparently clean import can halve the quantity or double the cost per item.

Use a staging area before updating the library

Prepare proposed records in a review table before writing them into a live system. Show whether each row is a new product, an update or an unresolved match. Search existing records by vendor and SKU before creating anything; use title and variant as supporting clues when an old record lacks a SKU.

Make exceptions easy to see: missing dimensions, conflicting currencies, unavailable images, unclear pack sizes, incomplete variants and products with no confirmed price. Leave unsupported details blank. A fluent description must not invent materials, durability claims or care instructions absent from the source.

Choose which images and descriptions the studio is permitted to reuse. Preserve source references so someone can recheck a specification later. Separate permission to access a trade source from permission to publish its information in a public catalog.

Verify the destination after the import

A successful import message is only one checkpoint. Read the resulting records back and verify the SKU, variant, price fields, images and quantity. Where a store is involved, also check publication and shipping settings; where a design schedule is involved, check which information a client can see.

Start with a small batch from one familiar vendor and include an existing item, a new variant and an incomplete record. Keep a before-and-after record of changes so mistakes can be corrected. Expand only after the rules handle those cases predictably.

1. Select one vendor and one destination, such as a product library.
2. Agree on the authoritative source for each important field.
3. Review a small proposed batch and resolve identity and pricing exceptions.
4. Apply the approved changes, read them back and update the vendor rule sheet.