Why ERP failed small businesses for thirty years

Small businesses are the largest category of business on earth and the last to be served by business software. For thirty years, ERP vendors looked at that market, saw the size of it, and tried to reach it. They mostly failed.

The usual explanation is price. It is wrong, or at least it is the shallow half of the answer. Vendors cut prices repeatedly — express editions, small-business editions, per-user cloud pricing — and businesses still did not adopt. Something else was going on.

The evidence that price was not the problem

Look at what vendors actually tried.

They shipped cut-down editions: the same product with fewer modules and a smaller price. They shipped preconfigured verticals: "ERP for distributors, ready in 30 days". They moved to cloud subscriptions, removing the capital cost entirely. They built partner channels so smaller resellers could serve smaller customers.

Each of these was a serious, well-funded attempt. Some found niches. None cracked the market. Meanwhile the spreadsheet — free, unsupported, actively bad at this job — kept winning, decade after decade.

When a free-ish product with terrible fundamentals repeatedly beats a well-engineered one, the well-engineered one is solving the wrong problem.

The real reason: the implementation cost does not scale down

Here is the structural fact everything else follows from.

An ERP implementation has a fixed cost floor that has almost nothing to do with your size. Somebody has to sit with you and find out how you work. Somebody has to translate that into configuration. Somebody has to migrate your data and train your people. That work takes weeks whether you have 20 staff or 2,000, because it is proportional to how many distinct processes you have — not to headcount.

A 20-person distributor has roughly the same number of distinct processes as a 200-person one. Fewer people doing each, but the same list: quote, order, pick, ship, invoice, chase, receive, pay, count. The discovery conversation is nearly identical.

So implementation lands as a near-fixed cost — and against a small company's budget it is enormous, often several times the software itself. The licence was never the barrier. The consultant was. Halving the licence price on a project where the licence is 20% of the cost changes nothing.

This is why every cheap edition failed. They discounted the part that was not the problem.

The second reason: cut-down editions removed the wrong things

When vendors built small-business editions, they had to decide what to cut. Reasonably, they cut the things big companies use and small ones supposedly do not: multi-entity, multi-currency, advanced costing, deep compliance.

But they kept the fixed data model, because that was the product.

So the business got a product that was still rigid, still needed configuring to fit, still required a consultant to bend — but now with fewer capabilities. They kept the expensive part and removed the valuable part. From the customer's seat: pay a consultant for weeks to install something that still does not fit, and has less in it.

The spreadsheet, whatever its many sins, fits perfectly. Because you built it, in an afternoon, exactly around how you work.

The third reason: businesses are less standard, not more

This one is counterintuitive and it is where most vendor strategy goes wrong.

The assumption behind every vertical edition was: small businesses are simple, so we can preconfigure best practice and skip the discovery.

But small businesses are not simple. They are idiosyncratic. A large corporation has been through consultants, auditors and a merger; its processes have been standardised toward the industry mean by force. A 15-person family distributor has thirty years of accumulated specificity — a payment term for one customer because of something that happened in 2011, an approval rule that exists because of one bad incident, a workflow shaped around one person's genuine expertise.

Those quirks are not inefficiency to be optimised away. Frequently they are the business. The reason that distributor still wins against a national competitor is precisely the thing no vertical template contains.

So "ERP for distributors, ready in 30 days" met a customer whose distinctiveness was their margin, and asked them to abandon it for someone else's best practice. Most sensibly declined.

What businesses did instead

Not nothing. They built a stack:

• Accounting software for the ledger, because that part genuinely is standard and the packages are good and cheap.
• Spreadsheets for everything else — the twenty processes surrounding the ledger.
• WhatsApp as the approval workflow.
• A shared drive as the document management system.
• One person's memory as the business-rules engine. Usually the longest-serving admin. Everyone knows who she is. Nobody has written any of it down.

This stack is much maligned and it is worth being honest: it works. It is flexible, everyone knows how to use it, it costs almost nothing, and it fits the business exactly.

It fails on scale, not on fit. It breaks when two people need the same row, when the person who knows the rules is on leave, when you cannot answer "how much did we sell last month" without an afternoon of work, and when nobody can tell you which of the four versions of the file is real.

Note what that failure list has in common: none of it is solved by better accounting software. That was never the gap.

What actually changed

The fixed cost floor was implementation. Implementation was expensive because translation was expensive. Translation was expensive because it needed a scarce bilingual human.

That is the specific thing that broke. Not licence pricing — the translation layer, as covered in natural language replaced the ERP consultant.

When you can describe a process and get working software, the fixed cost floor drops toward zero. And when the floor drops, the economics that locked businesses out for thirty years stop applying — not gradually, but structurally.

The other half is that idiosyncrasy stops being a problem. Under the old model, being unusual was expensive: every deviation from the vendor's model was a change request. Under a model where the system is generated from your description, your quirks are just what you described. The thing that made businesses unservable becomes a non-event.

That is the whole shift, and it explains why this era might reach the market that five previous eras could not. Not because software got cheaper. Because fitting got cheaper.

The honest caveat

None of this means every business should now buy an ERP.

If your accounting package plus a couple of spreadsheets genuinely works, keep it. The failure mode of this new era is buying software you did not need because it finally became affordable. Affordable and necessary are different things.

The businesses this matters for are the ones with a specific, recognisable symptom: you have outgrown the spreadsheet but a traditional ERP is a sledgehammer. You know the shape of it — the month-end scramble, the WhatsApp approvals, the person whose leave is a business risk, the question you cannot answer without an afternoon.

That gap sat unserved for thirty years, not because nobody noticed it, but because the economics made it impossible to serve. The economics changed. That is all — and it is enough.

Common questions

Why didn't cheaper ERP editions work for small businesses?

They discounted the part that was not the problem. An implementation has a fixed cost floor with almost nothing to do with your size: somebody has to find out how you work, translate that into configuration, migrate your data and train your people. That effort is proportional to how many distinct processes you have, not to headcount, and a 20-person distributor has roughly the same process list as a 200-person one. The licence was never the barrier.

Why do spreadsheets keep beating proper business software?

Because they fit. You built the spreadsheet in an afternoon, exactly around how you work, and it costs almost nothing while everyone already knows how to use it. The wider stack — accounting software for the ledger, spreadsheets for the twenty processes around it, WhatsApp as the approval workflow, a shared drive for documents, one person's memory as the rules engine — genuinely works. It fails on scale, not on fit.

Are my business's quirks a reason an ERP will never fit?

They were, under the old model, where every deviation from the vendor's data model was a priced change request. Vertical editions assumed small businesses are simple; they are not, they are idiosyncratic — a payment term for one customer because of something that happened years ago, an approval rule that exists because of one bad incident. Those quirks are often the business. When the system is generated from your description, they are simply what you described.

Should every small business now buy an ERP?

No. If your accounting package plus a couple of spreadsheets genuinely works, keep it — the failure mode of this era is buying software you did not need because it finally became affordable, and affordable and necessary are different things. The businesses this matters for have a recognisable symptom: outgrown the spreadsheet, but a traditional ERP is a sledgehammer. The month-end scramble, the WhatsApp approvals, the person whose leave is a business risk.

Background: how ERP actually evolved. If you are wondering whether you are in that gap, the fit check is built to tell you no when the answer is no.