A manufacturer’s operations had to move off a decade of Tally, Excel and paper registers onto a live ERP — without stopping the factory, losing history, or breaking the CA’s statutory world. Staged migration: stores first, then production, then costing, with parallel runs throughout and registers retired module by module.
Who this story is for
Any mid-size manufacturer where stock truth lives in a register, job-work lives in memory, and the annual costing exercise is archaeology — and where every previous attempt to change that stalled because the factory cannot pause.
The situation
A decade of accumulated process had produced a system that worked, as long as three specific people never took leave in the same week. Stores ran on a register. Job-work challans were tracked by the person who issued them. Costing was an annual exercise that produced a number nobody entirely trusted but everybody used.
The trigger was the usual one — [STORY SLOT: the specific event that forced it, in two lines] — and the constraint was absolute: the factory could not stop, and neither could the statutory accounting the CA had been running for years.
What I did
Sequenced by pain-to-risk ratio rather than by module completeness.
Stores and purchase first. The fastest win and the lowest risk: goods-received notes and issues went digital within weeks, with the paper register running in parallel until physical counts agreed three times running. Nobody was asked to trust the new system before it had earned it.
Production second, once material truth was stable. Work orders modelled on the floor’s actual flow rather than on the theoretical routing, with supervisor tablets replacing the whiteboard-photograph ritual. The design constraint here was hands that run machines: big targets, few decisions, forgiving forms.
Costing last, when real data was flowing. This ordering is not negotiable — costing built on unreliable stores data is fiction with decimals, and producing a confident wrong cost sheet early would have destroyed trust in everything else.
The Tally bridge throughout. Clean vouchers flowed to statutory accounting, so the CA’s world was fed better data rather than replaced. Winning that agreement early is a deliberate step: a CA veto arriving in month three has ended more factory ERP rollouts than any technical failure, and the objection is usually reasonable — they are protecting a statutory obligation nobody else in the room carries.
The result
The factory never stopped. Registers retired module by module, each one after its digital replacement had matched it repeatedly rather than on a scheduled date. And the first honest per-batch cost sheet in the company’s history arrived within the quarter.
That last one was uncomfortable, which is how you know it was real: at least one product everyone believed was profitable turned out not to be. Pricing built on actual costs is among the most underrated margin levers in manufacturing, and it is unavailable to any business whose costing is annual and approximate.
[METRIC — sign-off: stock accuracy before/after, month-end close time]
What it means for you
ERP migration horror stories are sequencing failures rather than software failures. The pattern that fails is big-bang: every module at once, on a date chosen by a project plan, with the old system switched off to force adoption. The pattern that works is staged, parallel-run, and retired by evidence.
The same discipline applied to a financial ledger is the zero-downtime migration, and the method is written up in full at /blog/zero-downtime-migration. The productised version of the factory system is Manufacturing ERP.
The three rules that made it survivable
Retire by evidence, never by date. Each register stayed until its digital replacement had matched it repeatedly. That takes longer than a cutover schedule and it removes the argument entirely — nobody has to be persuaded that the new system works when they have watched it agree for three weeks.
The floor is the customer. An ERP the shop floor routes around is an expensive report generator. Capture was designed for supervisors’ hands, in their language, at their pace, and the first version was corrected twice in the first month based on watching rather than on feedback forms.
Win the CA early. Statutory accounting stayed in Tally, fed clean vouchers. The accountant became an ally rather than an obstacle, because their obligation was protected rather than disrupted.
None of those three is technical, and together they are most of why this landed when previous attempts had not. The software was never the hard part; the sequencing and the politics were, and no vendor demo has ever covered either.
Related reading
/products/manufacturing-erp — the deployable build · /work/migration — the same method on live money · /blog/zero-downtime-migration — the recipe · /services/delivery — having it run for you
Factory chaalu, system badla
If every previous attempt died at “first we’d have to stop production” — that constraint is a solved problem with a sequence. /contact — barobar sequence se hoga.