Marketplace automation in 2026: stock sync, repricers and marketplace APIs — what to automate first and what it costs

What should a marketplace seller automate first?

Marketplace automation means stock, prices, orders and reports move between your seller accounts and your own records without anyone retyping them. Stock goes first because a mistake there costs the most: the last unit sells twice and one of the two orders has to be cancelled.

Automate what repeats, has clear inputs and an obvious action on failure. If a rule is argued over every time, such as how to price a new product, write it down before shopping for a tool.

How do you sync stock across Amazon, other marketplaces and your own site?

Stock is synchronised through one system of record: inventory software or a warehouse system that knows the real quantity and pushes it to every channel. Marketplaces do not share stock with each other, so without that hub a sale on one channel leaves the others showing an out-of-date number.

Stock in Amazon's fulfilment network is counted by Amazon and needs reports, not synchronisation; the problem starts when one shelf of your own feeds several storefronts.

Every sync has a delay, so on fast-moving lines sellers show one or two units fewer than the shelf holds. The buffer is cheaper than a cancellation: Ozon, for one, records an out-of-stock cancellation against the seller and tracks the cancellation rate in its seller rating.

What do the Amazon, Wildberries and Ozon APIs allow?

Each API has its own access model, batch size and speed limit, so an integration is designed around the strictest. VITON¹³ also serves sellers on Russian marketplaces, hence Wildberries and Ozon beside Amazon.

A throttled call returns 429, which Amazon calls retryable, though repeated throttling needs back-off. Amazon suggests the JSON feed from about 1,500 items; below that, the Listings Items API is faster within a five-minute window.

Do you need a repricer, and how do you set one up safely?

A repricer changes prices by rule: following competitors, holding a recommended retail price or reacting to stock. It earns its fee with hundreds of SKUs and rivals who reprice daily; for twenty SKUs a spreadsheet rule is cheaper.

• A floor from unit economics, or the algorithm follows a discounter into a loss; our unit economics guide shows the break-even price.
• Platform limits: Ozon allows 10 price changes per product an hour and rejects a minimum below 50% of your price.
• Wildberries quarantine: below a price threshold, the item keeps selling at the old price until you fix it.
• Native tools: Mercado Libre and Ozon have their own competitor-based pricing rules, and Mercado Libre blocks API price edits on items using its automation.
• Typo protection: the tool should refuse any price under the floor and alert someone.

Published repricer prices give a sense of scale: Informed lists a Launch plan at $99 a month for sellers under $5k in monthly revenue, Professional at $199 and Enterprise from $499.

Ready-made software or a custom API integration: which should you choose?

Choose ready-made software when processes are standard and speed matters, and a custom integration when your rules, systems or volumes do not fit. With software the vendor tracks API changes; with your own integration, you or your contractor do.

What breaks, and why do you need a log and an exception queue?

Code rarely breaks; the agreement between systems does: a key expires, a method changes, an unknown value arrives. The dangerous failures are silent: the marketplace says 'success' and nothing changes.

• Silent success: the Wildberries stock method does not validate parameter names, so a misspelt field returns 204 and nothing updates.
• Ignored fields: since 18 March 2026, Mercado Libre answers 200 OK to an update carrying a price for an auto-priced listing, but drops the price with a warning.
• Expired access: Wildberries tokens last 180 days, Ozon keys 3 months, and Ozon may revoke a key early over suspicious activity.
• Throttling: all three return 429 when you go too fast, and Ozon may restrict access without warning after a stream of faulty requests.

An exception queue lists the cases software will not decide alone: an unknown SKU, a price under the floor, negative stock. Each has an owner and a deadline; otherwise errors pile up in unread logs until the first wave of cancellations.

How do you roll out marketplace automation step by step?

1. List the manual workWrite down what is done by hand every day, where its data comes from and who owns the result.
2. Pick the system of recordDecide which system owns stock, price and product data. Two systems that both think they are in charge cause most mismatches.
3. Map products and warehousesLink your SKU to each marketplace's own identifier and your warehouses to theirs. A mapping error ships the wrong product.
4. Run a pilotConnect one category or one warehouse, reconcile your records against each channel every day for two weeks and record every mismatch.
5. Switch on the log and the queueKeep every request and response, send human decisions to the queue and diary every key expiry date.
6. Write the recovery planDescribe the fallback to manual work: who exports stock, how to pause the sync and how to catch up on missed orders.

How much does marketplace automation cost?

The cost has three parts: tool subscriptions, one-off design and build work, and upkeep after launch. Subscriptions recur, so compare them with a one-off build over a full year.

A worked example: Mirakl Connect Growth at $699 a month is $8,388 a year before add-ons, while Veeqo's inventory plan starts at $19 a month. A custom integration is a one-off build plus hosting and upkeep; your processes, not the price list, decide which is cheaper.