Off Tally, onto a real ledger: a 3-plant manufacturing ERP

Vishwa Precision Components machines and heat-treats safety-critical parts for automotive tier-1 buyers, out of three plants — Bhiwadi, Pune and Coimbatore. ₹180 crore turnover, nine years of history, and a system of record consisting of Tally ERP 9 plus roughly forty Excel workbooks with names like Stock_Master_FINAL_v3_USE_THIS.xlsx.

The problem was not that Tally was bad. It was that Tally was being asked to be an ERP.

Tally is excellent accounting software. It is not a manufacturing execution system, and it was never going to answer the four questions that were actually hurting this business:

• Where is my work-in-progress? Material issued to the shop floor left the books and reappeared, days later, as finished goods. In between, ₹6–8 crore of inventory was invisible. Plant heads estimated WIP. The word they used was "estimate".
• Which parts came from which heat lot? Their IATF 16949 obligations to OEM customers require lot-level traceability — if a batch of steel turns out to be bad, they must identify every part made from it. Answering that question took two people about two days of digging through paper job cards. In a live recall, two days is not an answer.
• What did we own on 31 March? Stock quantities lived in mutable Excel cells. There was no way to reconstruct a historical position, which meant every year-end valuation was an argument with the auditors.
• What is at the subcontractor? Job-work challans and the ITC-04 return were assembled by hand each quarter from a folder of paper. Material sitting at a job-worker was, functionally, material that had left the universe.

Month-end close took fourteen days. By the time the numbers were ready, they described a month that had stopped being actionable a fortnight ago.

Cutover was on 1 October, at the end of the second parallel quarter. It was uneventful, which is the highest praise an ERP cutover can receive.

The numbers

• Month-end close: 14 days → 3 days. The finance team did not become faster typists. The data stopped needing to be assembled, because it was already a ledger.
• Stock accuracy: 91.4% → 99.2%, measured by cycle-count variance against the system position. The remaining 0.8% is almost entirely a known shrinkage pattern at one plant, which they now know about because they can finally see it.
• Heat-lot traceability: 2 days → under 4 minutes. This one was tested for real, seven months after go-live, when an OEM customer raised a material query. They answered the same afternoon. Their quality head told us it was the first time in nine years that a traceability request had not ruined a week.
• 2.4M vouchers migrated with zero balance drift, verified against Tally across two full quarters of parallel running before anyone was allowed to switch anything off.

What it unlocked

WIP is now visible in near-real time, and the ₹6–8 crore that used to disappear between issue and finished-goods is a number on a dashboard. The ITC-04 job-work return, which used to be a quarterly two-person ordeal, is generated from the move ledger in one click, because material at a subcontractor is just a location in the ledger like any other.

The plant heads stopped estimating. That is the whole project, in three words.