Build vs. buy for small businesses: buy, connect, or build?

Three options: standard software, integrations, and a small custom app

It is rarely a simple build or buy. A company with 10 to 100 employees almost always needs all three, for different parts of the work.

The mistake we see most often: bending a standard package so far that it imitates a unique process. Or the reverse: having custom software built for something Exact simply does.

Decision rules

Ask four questions in this order:

1. How unique is the process? Does every business in your industry do it roughly this way? Then software for it exists. Do you really do it differently, because it sets you apart? Then it is a candidate for custom work.
2. How many people use it? One person with a spreadsheet: leave it for now. Four or more people working in the same file at once, emailing versions, and breaking each other’s formulas: that is an app.
3. How often does it change? Rules that change every quarter (price tiers, rates) belong in the settings of an app or package, not hardcoded. A process that has been the same for five years you can have built with confidence.
4. What does an error cost? A wrong price on 12,000 items, an invoice that never goes out, a missed deadline. The more expensive the error, the more a controlled app with logging and roles pays off.

Does a process score on unique, several users, stable, and expensive errors? Then building is defensible. In every other case: buy or connect.

The hidden cost of free Excel

Excel is not free. It looks free because it is already on every laptop. These are the costs that never show up on an invoice:

• One person understands it. When that person is sick or leaves, the process stops. In one of our cases, a single buyer managed the prices of 12,000 items in one file.
• Versions. “Masterlist_final_v3_NEW.xlsx” sits in three places and nobody knows which one is right.
• No log. Who changed that price, and when? Nobody knows.
• Errors your clients find. A formula that stops one row short. A copy-paste mistake. The client sees it on the invoice.
• No integration. The file is exported, imported, and exported again. Every step is manual work and a chance for an error.

Add up a year of hours and errors. It is often more than building a small app that does the same thing with roles, a log, and an integration.

Ownership of code and data

This is the part most often skipped in a proposal, and the part that hurts most when a supplier leaves.

With standard software, the data is yours, the software is not. Ask: can I export my data in a usable format if I cancel? And: what happens to my data if the supplier is acquired?

With integrations, you want to know where the workflows run (on your account or the builder’s), whether you can get to them, and whether the documentation is complete.

With custom software, our starting point is that the source code, the documentation, and the data are yours. No license on your own app. We put that in writing before we start. Ask every builder the same question, and be wary of a “no.”

Maintenance: who responds

Software that runs needs attention. A supplier changes its API, an AI model gets a new version, an employee comes up with a new exception. The question is not whether something changes, but who picks it up.

For integrations and custom apps the most important agreement is this: who watches it, and how do you notice an error? You can do that yourself if you have someone, or outsource it as Managed AI. An automation without maintenance is Excel with extra steps.

The choice as a process

As a diagram, the old situation is Supplier list comes in → Master list in Excel (1 user, no log) → Prices in ERP and online store, with the ring around the Excel. The new situation: Supplier list comes in → Normalize and check (integration) → Small app with 4 users and a log → Prices in ERP and online store. The ERP sits in the same place in both. Only the part in between has changed.

Example

A hospitality wholesaler with 28 employees ran its prices on one Excel file that one buyer managed. Replacing the standard software was not needed; the ERP worked fine. The solution was connecting (normalizing price lists automatically) plus a small app to replace the Excel, with four users and a log. An accounting firm with 18 employees did the same with its Excel tracker for documents clients send in. In both cases the client owns the code. Figures are illustrative. Read the cases price lists and backorders and document collection.