ERP vs off-the-shelf: how to actually decide

Most articles on this question are written by people who sell one of the two answers. We build custom ERP systems, so treat the following with appropriate suspicion — but we also talk clients out of them regularly, and the reasoning below is the same reasoning we use in that conversation.

Start by pricing the pain, not the software

The wrong first question is "what does an ERP cost". The right one is "what is the current process costing us", and it is usually answerable in an afternoon.

Count the hours. Someone re-keys sales orders into the accounts system. Someone reconciles three stock spreadsheets on a Monday. Someone spends two days a month assembling a report that leadership then does not trust. Multiply by loaded salary cost and you have a number.

Then count the errors that number hides: the stock-outs caused by a stale figure, the duplicate purchase because nobody could see existing inventory, the invoice raised late because it sat in an inbox. Those are usually larger than the labour cost and are almost never counted.

If that total is small, stop. Neither option is worth it, and a better spreadsheet plus some discipline will beat both.

The 90% test

If an off-the-shelf platform covers 90% or more of what you need, buy it. Not 70%. Ninety.

This threshold matters because the last 10% is where the cost lives. Closing a 30% gap through customisation, integrations and workarounds routinely costs more than the custom build would have, and you still pay the licence every year. You end up with the worst of both: someone else's data model, your maintenance burden.

Be honest when applying the test. "We could change our process to fit" is a legitimate answer sometimes — plenty of businesses have processes that are accidents of history rather than deliberate advantage. But if the process is the advantage, changing it to suit a vendor is a strange way to compete.

When buying is clearly right

• Your processes are genuinely standard. Straightforward distribution, conventional accounting, nothing unusual about how you fulfil. A mature platform has already solved this and solved it better than a first version of anything custom.
• You need it live in weeks. Custom does not go from nothing to organisation-wide in six weeks. An implementation can.
• Nobody will own it internally. Custom software needs someone on your side who cares about it. If that person does not exist, you are buying an orphan.
• The requirement is heavily regulated and standardised. Statutory reporting formats that every business files identically are not where you want to be writing original code.

When building is clearly right

• The mismatch is a daily tax. Not an annoyance — a measurable cost you can put a number on, every day, forever.
• You already pay for three systems that do not talk. The integration layer is often the bulk of the value, and once you are building that anyway the calculation shifts.
• Your process is your edge. If how you schedule, price or fulfil is why customers choose you, encoding it in generic software throws it away.
• Per-seat pricing has become absurd. At a certain headcount, licence renewal alone exceeds what a bespoke system would cost to build and run.

The middle answer nobody sells

You do not have to choose globally. The most common good outcome we see is buying the commodity layer — accounting, payroll — and building only the piece that is genuinely yours, with proper integration between them.

That keeps the custom surface small, which keeps it maintainable. A focused system that does three things exceptionally will outlive an everything-under-one-roof build that nobody wants to touch.

What to do this week

1. Put a number on the current cost. Hours plus errors.
2. Run the 90% test against two real products, with a demo using your data.
3. Get a scoped estimate for building only the part that failed the test.
4. Compare over five years, including licence renewals, implementation fees and internal time on both sides.

If the numbers come out close, buy. Custom only wins when it wins clearly, because it carries a maintenance obligation that a licence does not.