What Breaks When You Automate 250+ Marketplace Orders a Day: 6 Lessons From Our First Two Weeks
Our Unicommerce order automation now dispatches 250+ orders on busy days across six marketplaces. Getting there meant a slow panel, a marketplace that changed its process overnight, and one bug of our own. Here's what broke and how we fixed each one.
Two weeks ago we wrote about automating Unicommerce order processing for a multi-marketplace eyewear and accessories brand. Since then the system has gone from dozens of orders a day to 250+ on busy days, added Amazon Vendor Central, and moved to a tighter schedule that starts at 7:30 AM. Real volume found problems no test did. Here are the six that taught us the most.
1. One big batch is a single point of failure
On the first heavy day, 79 Flipkart shipments went into one manifest request. The request timed out after 90 seconds — but Unicommerce had actually created the manifest. Our run crashed, thinking it had failed. Now manifests go out in chunks of 25, timeouts are followed by a check of what really happened, and half-finished manifests are closed automatically. Any write that times out is never blindly retried — we ask the system what state things are in first.
2. A marketplace can change its process overnight, with no error
One morning every Flipkart invoice request simply hung — no error, no response, for minutes per order. The cause: Flipkart had started requiring orders to be accepted in its own panel before releasing labels, and Unicommerce's API had no way to say 'not ready yet'. We added a circuit breaker: two non-answers in a row on a channel and that channel is skipped for this run, so one stuck marketplace can't hold up the others. Those orders are retried next run and escalated to the team if they keep failing.
3. Escalate by persistence, not by count
Early alerts listed every hiccup — a throttled request that succeeded on retry looked like a problem. Now an issue is yellow the first time and turns red only if the same order is still stuck after two or three runs, with plain-language next steps for the team ('release the hold in the panel', 'repackage the cancelled item'). The team gets fewer emails and acts on all of them.
4. People and scripts shouldn't work the same orders at the same time
During one run, a team member bulk-processed orders in the panel at the same moment the script was invoicing them. Dozens of 'invalid state' errors followed. We agreed a simple rule — the script owns orders at run times, and anything a person processes by hand, the person completes — and taught the script to recognise those orders and leave them alone with a single line in the report instead of 50 errors.
5. Your own patch can be the bug
One morning's run crashed after invoicing 101 orders, before printing labels. The cause was ours: a code change the previous evening had been inserted one line too early, so a log statement ended up in the wrong function. Nothing was lost — the ledger knew exactly where each order was, and a rerun finished the batch — but the client got their orders two hours late. Since then, every change gets an automatic static check before it's deployed, however small.
6. Long-running browsers interfere with each other
Some jobs on the same server use a real Chrome browser (the price tracker, the Vendor Central run). A library we used re-patched one shared driver file every time any browser started — so starting a second job crashed the first one mid-run. Each browser now runs from its own private copy, and a nightly clean-up stops anything that's been running far too long. The server has two CPU cores; we schedule browser-heavy jobs in the quiet hours.
What stayed constant
Every shipment's stage lives in a ledger, so a crash means 'resume', not 'start over'. Irreversible steps are recorded the instant they succeed. And every change goes live on one real order first. On its best morning so far, the system dispatched 188 Flipkart orders in under 25 minutes, before most of the team had arrived.
If your marketplace operations are growing faster than your team, HowAutomate builds and runs this kind of automation for Indian sellers — including the boring parts that keep it working at volume.
Want this done for your business? See how we implement it end-to-end.
E-Commerce AutomationFrequently Asked Questions
How many orders a day can Unicommerce automation handle?
Our system dispatches 250+ orders on busy days across six marketplaces, and once processed 188 Flipkart orders in under 25 minutes. The practical limit is usually the marketplace and Unicommerce response times, not the script — which is why requests are chunked and slow channels are skipped rather than allowed to block the rest.
Why do Flipkart orders get stuck in Unicommerce?
A common cause is Flipkart requiring orders to be accepted in its own panel before labels are released. Until then Unicommerce shows the shipment as not ready for printing, and invoicing can hang without an error. Accepting the orders (or enabling auto-accept) releases them.
What happens if an automated run crashes halfway?
If every shipment's stage is kept in a ledger and irreversible steps are recorded the moment they succeed, a crash just means the next run resumes where the last one stopped — nothing is invoiced or dispatched twice.
Can the team still process orders by hand?
Yes, but not at the same moment the automation runs. The rule that works: the script owns orders during run times, and any order a person processes by hand, that person completes. The automation recognises those orders and leaves them alone.
Amit Singh
Founder, HowAutomate — Data Engineering, AI Automation & Cloud Infrastructure
Amit has 6+ years of experience building data pipelines, AI agents, and automation systems for businesses across India and globally. He founded HowAutomate to make enterprise-grade automation accessible to growing businesses.
Get Weekly Automation Tips
Real scripts, workflows, and AI tips — straight to your inbox.
Want us to implement this for you?
Book a free 30-minute discovery call and we'll map out exactly how to apply this to your business.
More Articles
Building a Marketplace Price Tracker for Amazon, Flipkart, Myntra & Nykaa (What Actually Works in 2026)
Read more
Automating Amazon Vendor Central Direct Fulfillment: Invoices, Labels, Handover & Ship Confirmation
Read more
Exam Photo & Signature Size 2026: IBPS, NEET, JEE Main & SSC (Verified From Official Notices)
Read more