The duplicate-item problem nobody plans for
Most stores do not connect Shopify to an empty Zoho organization. The items are already there. Someone entered them for purchase orders, an accountant imported them from a spreadsheet, or a previous integration created them years ago. Shopify has the same products, but the two catalogs were never built to line up. The SKU in Zoho has a prefix the storefront does not use. The Zoho name is the supplier's description while the Shopify title is the marketing one. Half the barcodes are missing on one side.
Then you turn on product sync. Every Shopify product that cannot be recognized as an existing Zoho item becomes a new one. You end up with two items for the same physical product, stock split between them, and sales reported against the wrong one. Cleaning that up afterwards is slow, because the duplicates now carry transactions.
The fix is to link the two catalogs before anything is created. That is what the AI Match page in Zync is for.
What the AI Match page does
AI Match, titled Jev AI product matching in the app, compares your unmapped Shopify variants with the items in your connected Zoho Books or Zoho Inventory organization and proposes which Zoho item each variant most likely is. You review the proposals and save the ones that are right.
Two things are worth being clear about from the start:
- The scan is read-only on the Zoho side. It does not create or update Zoho items. It reads them so it has something to compare against.
- Nothing is saved without you. Each suggestion sits in a table until you select Save mapping on that row.
A saved mapping is a link, not a copy. From then on Zync treats that Shopify variant and that Zoho item as the same product, which is exactly what product sync, inventory sync, and order sync need in order to update the right record instead of creating another one.
How a scan works: exact checks first, then Jev AI
A scan looks only at variants that are not linked to a Zoho item yet, and only at Zoho items that are not already linked to something in your store. For each variant it works through the cheap, certain checks before asking AI anything:
- Exact barcode. If exactly one available Zoho item carries the variant's barcode, that is the suggestion.
- Exact SKU and name. If exactly one Zoho item has the same SKU and its name shares words with the Shopify title, that is the suggestion. The SKU alone is not enough, because reused SKUs are common.
- Exact title/name. If exactly one Zoho item has the same name as the product title plus variant title, and the SKUs do not contradict each other, that is the suggestion.
Only when none of those settle it does the variant go to Jev AI. Zync builds a shortlist of the most likely Zoho items for that variant and asks Jev to pick the one that is the same sellable product, or to answer that none of them is. A variant with no plausible candidates gets no suggestion at all, which is the right outcome. A blank is easy to deal with. A wrong link is not.
Each suggestion carries a reason so you can see which path produced it: exact barcode, exact SKU and name, exact title/name, or a Jev AI match.
Focused scan or whole store
There are two buttons on the page and they suit different moments.
Scan matches runs a focused scan. Type into the Search Shopify box first. It accepts any product or variant ID, SKU, barcode, name or title, vendor, or product type, and checks all of those fields automatically. Use it when you are fixing one product, working through one vendor's range, or testing how matching behaves before committing to the full catalog.
Scan whole store checks every unmapped variant. Use it once at setup, and again after a large import.
You can also start from the Products page. On the Product Sync tab, Match with Jev AI opens the page, and each unsynced product has its own AI match button that opens it with that product already in the search box.
Scans run in the background as queued jobs. The page shows progress as variants are processed, and you can leave and come back. One scan runs per shop at a time, and the results stay available for 24 hours.
Reading the suggestions table
When the scan completes, the Suggestions section lists one row per proposed match with these columns:
- Shopify title, Shopify SKU, and Shopify barcode for the variant.
- Zoho name, Zoho SKU, and Zoho barcode for the proposed item.
- Confidence, as a percentage.
- Reason, the check that produced the match.
- Action, with the Save mapping button.
Read it left to right as a comparison, not as a verdict. An exact barcode match with matching names needs a glance. A Jev AI match where the two SKUs differ deserves a real look at pack size, colour, and unit, because those are the details that separate two genuinely different items with similar names. Rows are sorted with the highest confidence first, so the top of the list goes quickly and your attention goes to the bottom.
If a scan finds nothing for a product, the page says so and suggests checking the item's Zoho name and SKU, or syncing the Shopify product to Zoho so a new item is created on purpose.
What happens when you save a mapping
Selecting Save mapping writes the Zoho item ID to the Shopify variant in the custom.zoho_item_id metafield. That metafield is the link every other sync reads.
What happens at the product level depends on the product:
- Single-variant products get the same link on the product as on the variant.
- Multi-variant products are linked one variant at a time. Once all variants are mapped to Zoho items that belong to the same Zoho item group, the product itself is linked to that group.
That second rule matters. A product with three of its four variants mapped is not treated as unsynced. If you sync it, Zync sees the existing variant links and does not create a fresh item group over the top of them. Finish the mapping, or map the last variant by hand in Zoho and scan again.
You can read more about how these links are stored and used in how records are linked.
The guardrails that keep matches safe
A matching tool is only useful if a mistake is hard to make. Several checks run when you save, not just when you scan:
- One Zoho item, one variant. The same Zoho item is proposed for only one Shopify variant in a scan, and a save is refused if that item has since been mapped elsewhere.
- Already-mapped items are off the table. A Zoho item linked to another product in your store is never suggested, even in a focused scan of a different product.
- Stale results are rejected. If you start a new scan, suggestions from the old one can no longer be saved. You are asked to run a fresh match first.
- Existing links are not overwritten. A variant that already points to a different Zoho item cannot be re-pointed from this page by accident.
The net effect is that the worst case of a wrong suggestion is a row you do not accept.
Where Product Auto-Matching fits
Zync has had a simpler matching option for a long time, and it still has its place. Under Products, on the Product Settings tab, Product Auto-Matching (Shopify to Zoho) lets you pick one field, SKU (Internal Reference), Barcode, or Product Name, and when a product syncs Zync looks for an existing Zoho item with the same value before creating a new one.
If your SKUs are clean and identical in both systems, auto-matching is all you need. Turn it on and sync.
AI Match is for the catalogs that are not that tidy. It looks at several fields together, it copes with names that are similar rather than identical, and above all it shows you the match before anything is linked. The two work well in sequence: map the awkward products on the AI Match page, then let auto-matching and normal sync handle the rest. The settings are covered in product settings.
A sensible order of work for a first sync
If you are connecting a store to a Zoho organization that already has items, this sequence avoids nearly all duplicate cleanup:
- Connect Zoho and pick the right organization. AI Match needs a connected organization to read items from.
- Leave product auto sync off for now, so nothing is created while you are still mapping.
- Run Scan whole store and accept the exact barcode and exact SKU matches after a quick check.
- Go through the Jev AI matches more slowly. Skip anything you are not sure about.
- For products with no suggestion, decide whether the Zoho item is missing or just named very differently. Fix the name or SKU in Zoho and run a focused scan, or accept that it will be created.
- Turn on Product Auto-Matching, then sync the remaining products from the Product Sync tab.
After that, inventory and orders have a clean set of links to work with. Order lines find their Zoho item through the same variant link, so a properly mapped catalog also means cleaner invoices. The post on variants and item groups explains what gets created for the products that do need new items.
What AI matching does not do
It helps to know the edges:
- It does not create, rename, merge, or delete Zoho items.
- It does not save anything on its own, whatever the confidence.
- It does not re-examine variants that are already linked. To change an existing link, unlink the product on the Product Sync tab first.
- It does not handle pack-size conversions. If one Shopify SKU should become 24 units of a Zoho item on an order, that is product mapping on orders, a different feature.
Next step
Open AI Match in the app, type one product you know exists in both systems into the search box, and select Scan matches. Seeing one correct suggestion with its reason is the fastest way to judge how your catalog will behave. The full walkthrough is in the AI product matching guide, and the feature overview is on the AI product matching page.