01/ 07

The prompt

Give it to Claude, ChatGPT or Cursor along with your key. It reads the docs, follows the recipe, and asks you before spending anything.

Recipe 2.6

Read https://data.magpieiq.com/llms.txt and the docs at https://data.magpieiq.com/docs.
Then fetch https://data.magpieiq.com/skill/SKILL.md and follow recipe 2.6.
My Magpie API key is: <paste your key>

Rules: submit only the URLs or items I give you; poll with backoff rather than tight-looping; write the raw payload to disk before analysing it, because re-running a job costs credits again; report anything that failed rather than quietly returning a shorter table; ask me before spending any credits.
This recipe is approval-gated and only processes 08:00-18:00 GMT+8, so do not build an unattended schedule on it.

What I want: for these eight Shopee products, tell me which variant actually sells. Rank the variants by units, give me each one's share of the product total, and say whether demand is concentrated or spread.

02/ 07

What it tracks

01

Sold count per variant, ranked

02

Each variant's share of the product's total sold count

03

Whether demand concentrates in one or two variants or spreads evenly across the range

04

Which submitted items returned variant data and which did not

03/ 07

How it works

1
Assemble your items as {item_id, shop_id, country}. One country per request, up to fifty thousand items.
2
Submit to variant_sold_v1 and poll. Approval-gated — hand back the job_id rather than polling.
3
Rank variants by sold count and compute each one's share of the product total.
4
Report whether demand concentrates or spreads, because that is what decides whether the range needs pruning.

04/ 07

What a real response looks like

Real variant_sold_v1 response · one Shopee listing · top 3 of 47 models by units sold

            GET /v1/shopee/variant_sold_v1/download/{job_id}
            captured 2026-08-28
          
{
  "data": {
    "items": [
      {
        "itemId": "19625281827",
        "name": "MAYBELLINE OFFICIAL Superstay Vinyl Ink 4.2 ml - Shiny Pigmented Liquid Lipstik Lipstick Make Up Lipcream Longlasting Waterproof Tahan Lama 16 Jam Lipstik Lipgloss Lipbalm Liptint Lip Glossy",
        "currency": "IDR",
        "models": [
          {
            "modelId": "211312317264",
            "name": "10 Lippy",
            "price": "10660000000",
            "sold": 237664,
            "stock": 222
          },
          {
            "modelId": "211312317265",
            "name": "60 Mischievous",
            "price": "11600000000",
            "sold": 176252,
            "stock": 96
          },
          {
            "modelId": "211312317259",
            "name": "65 Saucy",
            "price": "10660000000",
            "sold": 175512,
            "stock": 70
          }
        ]
      }
    ]
  }
}

05/ 07

Tips from our own runs

Send one country per request, since a job can't mix countries. Splitting them up keeps each result easy to read.

Submit during 08:00–18:00 GMT+8, when approvals are processed, and trigger it yourself rather than on a schedule.

If most sales land on two or three variants, that's a real finding worth showing, not noise to smooth out.

06/ 07

When not to use this

Use this when

Variant demand profiler

You know the Shopee listing and need sold counts per size, shade or pack.

Use instead →

Category share tracker

This profiles variants within products you already know about. To find which products to profile in the first place, start with the Category share tracker or the New-launch watch.

07/ 07

Other tools

Get a key, then paste the prompt.

New accounts start with free trial credits. You generate your own key from the dashboard.

Get an API key Read the recipe in the docs

See which variant actually sells.