When Does a Growing Indian Business Actually Need an ERP?

Ask a hundred consultants when a business needs an ERP and you will get thresholds: a turnover figure, a headcount, a branch count. These are proxies, and poor ones. The real signal is operational, and you can observe it this week. Pull the stock figure for your fastest-moving item from the godown register, from Tally, and from the sales team's spreadsheet. If you get three numbers, you have your answer.

The disagreement is not carelessness. Each record is locally correct — the godown keeper wrote down what physically moved, the accountant entered what was invoiced, the salesperson noted what was promised. They diverge because they are three separate copies of what should be one fact, updated by different people at different times, and reconciled by a human being whenever someone notices. Reconciliation effort grows with transaction volume; human attention does not. Past a certain size, the copies drift apart faster than anyone can pull them back together.

The visible cost is the labour — the person who spends the first week of every month assembling the GST filing by hand from sources that never quite match. The invisible cost is worse: decisions made on numbers that are wrong. Committing stock you do not have. Extending credit to a customer whose real outstanding is higher than the sheet shows. Reordering material that is sitting, uncounted, in the second godown. When the numbers stop agreeing, the business is navigating on a map that no longer matches the road.

Strip the jargon and an ERP is one thing: a single source of record. One database where a customer's order, the stock it reserves, the invoice it becomes and the ledger entry it creates are the same fact — written once, at the moment it happens, by the person it happens to — and merely viewed from different screens. The sales screen, the godown screen and the accounts screen are windows onto one dataset, so they cannot disagree, in the same way two windows of the same room cannot show different furniture.

That is the whole trick, and it is also why the category has a reputation for both transformation and trauma. Because here is what an ERP is not: it is not a reorganisation of your business. It does not decide who may approve a discount, what happens when a customer wants goods before payment clears, or which godown ships first. It records the process you already have — and if 'the process' is actually five processes, one per senior employee, the software surfaces that on day one. Every question your business has been fudging verbally becomes a mandatory field.

This is worth stating plainly because it is the root of most ERP disappointment. Businesses buy the software expecting it to impose order, and discover it instead exposes disorder — precisely, publicly and immediately. Whether that is a gift or a catastrophe depends entirely on whether you were ready to answer the questions.

Any honest discussion of ERPs in India has to start by giving Tally its due. It is ubiquitous for good reason: your accountant knows it, your auditor expects it, and for the books of account and GST compliance it does its job well. Businesses do not outgrow Tally's accounting; they outgrow everything they have bolted on around it.

Look at what actually surrounds Tally in a typical trading or distribution firm: an Excel sheet for stock, because operations needs the numbers cut a way the accounts entry does not capture — by batch, by godown, reserved versus free; a WhatsApp group where orders get 'confirmed'; a phone call or a notebook for approvals; another spreadsheet reconciling stock across branches. That ring of spreadsheets and chats is where the seams split, because every sheet is a private copy of the truth — owned by one person, updated when they remember, shared as an attachment that is out of date the moment it is sent.

Notice the pattern: Tally sits at the end of the pipeline, faithfully recording outcomes — but the pipeline itself, from enquiry to order to dispatch to invoice, lives in tools that were never designed to hold a workflow. The fix is rarely 'replace Tally'. In most builds we scope, the sensible architecture keeps Tally as the books of account and puts a proper operational system in front of it — one that owns orders, stock and invoicing, and hands the accountant clean, consistent data instead of a shoebox of sheets to reconcile.

The classic ERP failure has a recognisable shape: months of configuration, every department's process rebuilt at once, all staff retrained in the same fortnight, and a single go-live day on which the entire business switches over. Go-live becomes the first day the system meets reality — the highest-risk day of the whole project scheduled, by design, at the point of maximum commitment. When something breaks, everything breaks together, and the fallback is the old spreadsheets, which quietly become the real system again while the new one gathers dust.

The alternative is to implement one painful flow first — completely, honestly, in production — and let it earn the right to expand. For most Indian trading and distribution businesses, the right first slice is orders to invoice: an order captured once, stock checked against a number the system actually controls, and a GST invoice generated with serial numbers that never skip and never repeat — which, since GST invoices must be serially numbered, is a compliance property as much as a tidiness one. That single flow removes the WhatsApp-to-invoice relay, and it produces a visible result the whole team can judge within weeks rather than quarters.

This is how we build our own products at NEXINFINITY META, not just client systems: ship one working vertical slice, run it against real transactions, then extend to the next — stock, then purchases, then approvals, then branch views. Each slice goes live while the rest of the business carries on untouched, so the blast radius of any mistake is one flow, not the whole company. By the time the later slices arrive, the team already trusts the system, because they watched the first one work.

Custom software has a reputation for open-ended cost, so here are our real numbers. At NEXINFINITY META, a well-scoped slice — roughly a month of senior build time — runs about ₹1.5–2.5 lakh (roughly $2,000–$3,500). A larger system — multi-branch, multi-godown, several connected flows — starts from several lakh ($5,000+). Every engagement is a fixed written quote against a written scope, and the client owns 100% of the code: no licence to renew, no vendor who can reprice you, no per-user meter running.

The economics work because AI now does the boilerplate — the forms, the data plumbing, the report scaffolding — while senior engineers spend their time on the parts that decide whether the system is actually right: the data model, the stock-movement rules, the serial-number guarantees, the permission boundaries. That is efficiency, not corner-cutting; the full picture of how we scope and price this kind of build is on our custom SaaS development page.

Compare the shape — not any particular vendor, the shape — of subscription ERP pricing. Per-user, per-month fees multiply with headcount and never stop: every hire raises the bill, every year restarts it, and after five years you have paid sixty months of rent and own nothing. That can still be the right choice for a business whose needs closely match an off-the-shelf product. But for a business whose operations are its edge — an unusual credit cycle, a specific approval chain, a branch structure the packaged products fight — the custom build stops being the expensive option surprisingly quickly, because its cost is a one-time build plus modest maintenance, while the subscription's cost is a function of your growth.

An ERP rewards businesses that are ready for it and punishes those that are not, so the last question is not really about software. Start with the disqualifiers. If your processes change weekly — if how an order gets approved depends on who happens to be in the office — software will not save you; it will simply make every change more expensive, because now every change is a software change. Fix the process on paper first, run it manually for a month, and then automate what survived.

Similarly, if the plan is for the owner to stay on WhatsApp while the staff use the system, stop. A system of record only works when it is the record — the moment senior people transact outside it, everyone learns that the real business happens elsewhere, and the spreadsheets return within a quarter. And if nobody can name the single most painful flow, the pain is probably not yet expensive enough to justify the change; wait until it is, because you will need that pain as fuel through the adoption weeks.

The ready signs are simpler — and if most of these describe you, the disagreeing numbers you noticed this week are not a nuisance to live with. They are the starting gun: