End-to-end visibility before the first peak
The state of every order lives in one place from launch, not after volume overwhelms the team. It was the partners' main reason to start before going live.
Manufacturing·Miistyc
From five tools that do not talk to each other to an order you can follow end to end.
How Miistyc connects its store, its payment gateway, its accounting, its carrier and its WhatsApp under an orchestrator that knows where every order stands, before volume grows.
5 systemsShopify, Mercado Pago, Siigo, Envía and Jelou
12 connectionsbuilt by forabi between the tools
7 statesper order, from created to delivered
The problem
Miistyc is a food brand (granola and similar products) with its own production plant that launched in September 2026 and sells 100% online: a Shopify store, traffic from social media and customer service over WhatsApp, with no physical storefront.
Before selling its first bag it already had five tools: Shopify for the store, Mercado Pago (a payment gateway) to collect payments, Siigo (Colombia's most widely used invoicing platform) to invoice, Envía to ship and Jelou for WhatsApp. Each does its job well and none knows what is happening in the others. Answering "where is my order?" meant opening three platforms, the shipping manifest was signed on paper and inventory was kept in a handwritten ledger.
The partners set one clear condition: the experience had to feel human, without bombarding the customer with duplicate notifications. And they wanted it solved before launch, not after volume overwhelmed the team:
- Every paid order had to be invoiced, booked and flagged to production and logistics by hand.
- The same customer question was answered by checking the store, the gateway and the carrier separately.
- Without a single source of truth, a repeated webhook or a reversed payment could be invoiced twice or not at all.
What we built
We built a headless order orchestrator: the team keeps working in Shopify while the orchestrator listens, queries and acts on the five tools. Every notification arrives by webhook with a verified signature, is deduplicated and moves the order through its state machine: created, payment confirmed, invoiced, in preparation, shipped, in transit, delivered.
- 01Signed intake with no duplicates: Shopify and Mercado Pago send notifications; the orchestrator verifies the signature, rejects anything tampered with and stores each event exactly once, even if the provider retries.
- 02Invoicing in Siigo when payment is confirmed: creates the customer record, issues the electronic invoice and records the payment receipt; a database constraint blocks a second invoice for the same order.
- 03Hourly monitoring: finds paid orders without an invoice, shipments with no updates and stuck notifications, and puts them in a human review queue instead of failing silently.
- 04The oven and the cameras, wired to the line: we connected the oven and the plant's camera system to measure how long each person took at each step of the production line; that data showed where the constraints and bottlenecks of the operation were.
- 05One channel for the customer: pickup, transit and delivery updates go out over WhatsApp through Jelou, one per state; the question "where is my order?" is answered from the database itself by order number, tracking number, email or phone.
The orchestrator already listens to Shopify and Mercado Pago with the history imported (real orders, payments and tracking numbers) and runs in shadow mode: it computes the state of each order without writing to Siigo or messaging the customer yet, until the team confirms over a week that what it computes matches reality. Invoicing, WhatsApp updates and the Jelou webhook are next.
The result
Business rules the database enforces
A second notification for the same state, a repeated webhook and a second invoice for the same order are rejected by the schema, not by a condition someone can forget.
What the real data taught us
Importing the history showed that no order carried a customer ID number and that tracking numbers came from four different carriers. Solving it in shadow mode avoided invoicing wrong from day one.
Bottlenecks measured, not guessed
Crossing the oven timings with the camera timings showed which steps slowed the line and who was carrying them. Plant decisions stopped depending on the shift's perception.
Honest scope
Inventory was left out by joint decision and reversals are handled with manual credit notes. It is 12 connections and 5 actions, not the 19 from the initial survey.
The Forabi Method
- 01Free consultation
Discovery with the partners and area leads; we mapped the 19 connections between the five tools and which ones already existed.
- 02Design
Order state machine, seven business rules and the edge cases settled in an in-person meeting: reversals, duplicates, carrier exceptions.
- 03Build
Functions on Supabase with business rules tested offline, signed webhooks, a cron processor, monitoring and order-status lookup.
- 04Shadow mode
One week comparing what the orchestrator computes against reality before letting it write to Siigo or message the customer.
- 05Ongoing support
Invoicing, WhatsApp updates and the later phases are next: subscriptions and good manufacturing practices on the plant floor.