Day 1: Buying from Multiple Retailers at Scale

Today we're launching multi-product, multi-quantity ordering across retailers. One API call to build a cart with multiple SKUs, set quantities, and check out.

Before this, every growing company hit the same wall: We need to buy a lot of stuff, from a lot of places, and doing it manually doesn't scale. Maybe it's a procurement team ordering supplies from fifteen different vendors. A fulfillment company sourcing products from wherever they're cheapest. A gifting platform sending curated packages from Ulta, Costco, and Amazon.

The options aren't great:

Do it yourself. Someone on your team spends hours clicking through checkout flows. It works until it doesn't. Order volume grows and that person becomes a full-time purchasing clerk.

Hire VAs. Cheaper per hour, but you're still paying for manual work. And manual work means manual errors: wrong quantities, wrong addresses, wrong items. You catch the mistakes in customer complaints. Plus, managing a team of VAs is its own headache. Training, time zones, QA, turnover. You traded one problem for three.

Use RPA. Record a bot clicking through a website. It works great until the retailer changes a button color or moves a field. Then it breaks silently and you find out when orders stop going through.

None of these solutions actually solve the problem. They just move it around.

What makes this hard

Most retailers don't have purchasing APIs. Not Amazon. Not Target. Not Walmart. Not anyone, really. At least not for programmatic purchasing at scale.

So if you want to automate this, you're interacting with websites built for humans, not machines.

Every retailer handles carts differently. Some persist cart state in the session. Some tie it to an account. Some have aggressive bot detection that triggers the moment you add a third item too quickly.

Then there's checkout. Multi-item carts mean more failure points. An item goes out of stock mid-checkout. A quantity limit kicks in. Shipping calculations change and suddenly your total doesn't match what you expected.

When you outsource this to VAs or RPA, these edge cases become your problem. The VA doesn't notice an item was substituted. The RPA bot doesn't know how to handle a "only 3 left in stock" warning. You find out after the fact, when something's wrong.

At scale, across dozens of retailers, this multiplies fast.

How we solved it

There are two ways to approach this problem.

Deterministic systems try to map every retailer's checkout flow explicitly. If button X appears, click it. If field Y exists, fill it. This works until it doesn't. Retailers change their sites constantly. A deterministic system is always playing catch-up, and it breaks in predictable ways when something changes.

Adaptable, computer-use models work differently. Instead of hard-coding every flow, you build a system that understands what it's trying to do (add items to a cart, enter shipping info, complete checkout) and figures out how to do that on whatever site it's on.

That's what we built. Zinc's ordering engine uses an adaptable model that can navigate checkout flows across retailers without needing explicit mappings for every site. When a retailer changes their UI, we don't need to update a script. The system adapts.

This is how we went from supporting multi-item orders on a handful of retailers to supporting them almost everywhere.

What this unlocks

One API call. Multiple SKUs. Multiple quantities. Works everywhere we support.

{
  "retailer": "walmart",
  "products": [
    {
      "url": "https://www.walmart.com/ip/12345678",
      "quantity": 2,
      "variant": [
        { "label": "Color", "value": "Navy" }
      ]
    },
    {
      "url": "https://www.walmart.com/ip/87654321",
      "quantity": 5
    },
    {
      "url": "https://www.walmart.com/ip/11223344",
      "quantity": 1
    }
  ],
  ...
}

Build a cart the way a human would, but programmatically, at scale.

No VAs. No RPA scripts to maintain. No manual errors showing up in customer complaints.

Try it for yourself!

Start a new order directly in the dashboard. Add items, set quantities, and test the full flow before writing any code.

For rules around quantities, variant selection, and handling partial failures, check out Multiple Products & Quantities in our docs.