Skip to content

Automation · Operations and reporting

Save the weekly Mercado Libre order report to Google Sheets

Scan one closed week of visible Mercado Libre Colombia orders and save a buyer-free aggregate by status to Sheets after explicit approval.

How do I save a verifiable weekly Mercado Libre order report to Google Sheets?

You run the task manually with the Monday of a week that has ended and one empty row on the Weekly Meli tab. Your Met generates a random UUID, queries one Mercado Libre Colombia account and walks no more than 20 pages or 1,000 candidate orders. It applies the exact interval from Monday 00:00 up to, but not including, the next Monday 00:00, and returns counts and order total_amount by status, units and the top 20 products by units. It returns no buyer, payment or order number. An incomplete scan never writes. A complete one displays the UUID, hash, exact range and all 13 values; only after your approval does it issue one RAW update and read the row back. This is an operations report, not net revenue, profit or settled money.

What starts it

When someone asks for it

A person runs the task with week_start, which must be the Monday of a week already closed in America/Bogota, the spreadsheet_id and one exact empty row between A2:M2 and A3500:M3500 on the Weekly Meli tab. There is no implicit cron or recurrence.

A manual task: someone on the team runs it with a button, or asks the Met for it in chat.

Who runs it

Met for marketplace operations

It has no seeded proper name. Its job is to close a bounded week without turning order data into accounting or copying buyer information to a spreadsheet. Approval separates the marketplace read from the single external write.

What you need connected

4 steps · 5 tools

The procedure, step by step

Each step shows the tool that runs. They belong to the real connector: if one of them stopped existing, this page would not build.

  1. 1

    Confirm the account and retain only the seller site

    The account query confirms that the credentials belong to the expected seller and site_id is MCO before requesting a week. The tool returns only the technical id, nickname, country_id and site_id; it does not expose a first name, last name or email. A different site or unexpected account stops the task. The weekly tool queries /users/me again and revalidates MCO: this first step lets the person spot a wrong connection before the scan and never replaces the technical gate.

  2. 2

    Close one Colombian week and prove the scan is complete

    Your Met creates an opaque random UUID v4 with no email, phone, name or order number, then calls the tool with the selected Monday. The authenticated account must be MCO and every amount must be COP. The query asks from Monday 00:00 to the following Monday 00:00, then applies the half-open interval locally: Sunday at 23:59:59 is in and the next Monday at 00:00 is out. It walks pages of 50, up to 20 pages and 1,000 orders. If paging.total changes, an ID is missing or repeated, a page ends early, another currency appears or the cap is exceeded, it returns complete:false. That result is shown as incomplete and never reaches Sheets.

  3. 3

    Check the tab, headers and one empty row

    Metadata must show the exact Weekly Meli tab. Columns A:M are operation_key, report_hash, week_start, week_end, timezone, site_id, currency_id, visible_order_count, item_units, order_amount_total, orders_by_status, amount_by_status and products_by_units_top_20. A bounded read selects one fully empty row from 2 through 3500. A missing tab, changed header or any existing value stops the task; it never falls back to another tab or approximate range.

  4. 4

    Show, wait for approval and write exactly once

    A complete result includes the UUID, a SHA-256 covering the key and full report, and buyer-free aggregates. Your Met displays that UUID, hash, selected A:M row and all 13 values in order, including the three JSON cells. It also displays the limits: total_amount by current status is not revenue; fees, withholding, shipping, returns and Mercado Pago reconciliation are absent; and the seller view does not expose every cancellation. Only explicit approval of that payload enables confirmar=true. Changing the week, range or any value invalidates approval and requires a new hash. After approval, the dedicated tool again requires Weekly Meli, UUID, hash, report and confirmar=true. It rereads A1:M1 and all 13 exact headers, then pre-reads the row and, only while empty, sends one RAW values.update with retry:false. It reads back all 13 cells and never resends an ambiguous update; only an exact match is success. Pre-read and update are not CAS, so another writer can still intervene and atomic uniqueness is never promised.

Before you start

What you need ready

The Met solves none of these for you. If one is missing, the automation stops there.

The limits

What this automation does NOT solve

Better said here than discovered halfway through the rollout.

FAQ

Questions about this automation

It adds total_amount from the orders returned by the seller view and groups them by current status. It does not subtract fees, withholding, shipping or returns, and it never queries a Mercado Pago settlement. Calling it net revenue would mix different contracts.

Yes. The local interval includes the starting Monday and all of Sunday up to, but not including, the next Monday at 00:00. Since the API reduces date-filter precision, the tool requests the boundary and filters each date_created locally again.

The tool returns complete:false before anything is written. Use a future tool that can narrow the period with the same guarantees or an asynchronous process designed for that volume; this recipe neither splits the week nor presents a sample as complete.

Yes, that race exists because read and update are two calls, not an atomic comparison. The tool reduces risk with an exact row, key, hash, pre-read, one update and read-back, but concurrency means inspecting the range and never retrying blindly.

Keep going

Other automations built on the same tools

Want this automation running on your own data?

We will show it to you with your accounts connected, not with a canned demo.