Picture a distribution business with two warehouses, about 1,500 products and 25 staff. Orders come in by phone, email and a web store. Stock lives in a spreadsheet that is right on Monday and wrong by Wednesday. Invoices are raised in the accounting package, purchase orders in a different spreadsheet, and the owner finds out about a stock-out when a customer complains.
At some point the owner starts looking at ERP for small business and finds three paths: buy a ready-made system, take a flexible one and customize it, or have one built. Each is the right answer for some companies and an expensive mistake for others.
Start with the processes, not the software
An ERP (enterprise resource planning system) is simply one database that several departments share: sales, purchasing, inventory, accounting and, for some businesses, manufacturing, HR or field service. Its value comes from people no longer re-typing the same information into different places.
Before you look at any product, write down the five processes that cost you money when they go wrong. For our distributor they might be:
- Taking an order and knowing, at that moment, whether the stock is really there.
- Reordering from suppliers before items run out.
- Moving stock between the two warehouses.
- Invoicing and chasing payment.
- Seeing margin by customer and by product.
For each one, note who does it today, which tools they use and where it breaks. That page is worth more in a vendor meeting than any feature checklist, because it lets you judge every option against your real work.
Buy: when an off-the-shelf ERP for small business is enough
If your processes look like most businesses in your sector, buy. Accounting, invoicing, standard inventory, purchasing and basic reporting are solved problems, and the established products handle them well. Odoo, ERPNext, Zoho's business apps, Microsoft Dynamics 365 Business Central and NetSuite are among the names you will come across, at quite different price points and levels of complexity.
What you gain: tax and compliance updates that the vendor usually maintains, a product other companies have already tested, and implementation measured in weeks or a few months rather than a year.
What you give up: you change your process to fit the software. That is often healthy, because many "special" processes are habits rather than advantages. But if the software forces a workaround into something your customers notice, such as how fast you quote or dispatch, that cost is real.
Buying is usually the wrong choice only when you have already tried one or two packages and your team is running the business from spreadsheets exported out of them.
Customize: a configurable ERP with your own modules
The middle path is a system designed to be extended. Two well-known examples are open source: Odoo's Community edition is released under the LGPLv3 license (its Enterprise edition is proprietary), and ERPNext is released under the GNU GPL v3. You can host either yourself or use a hosted service, and add your own modules for the parts of your business that do not fit.
This works well when most of your needs are standard and a smaller part is specific to you. The distributor might use standard accounting, purchasing and inventory, and add a custom module that allocates stock between warehouses using its own rules.
The discipline that makes it work: never edit the system's core code. Put every change in a separate module. Core edits make each upgrade a small rebuild, and teams that skip upgrades end up on an old version that is hard to support. Budget for testing your customizations at every major upgrade, because some rework is normal even when the modules are built properly.
Build: when a custom ERP earns its cost
Building your own makes sense in fewer cases than vendors of custom software like to admit. We would consider it when:
- The process is your advantage. A logistics company whose pricing, dispatch or route rules are how it wins work does not want those rules squeezed into someone else's screens. If that sounds like your business, see our page on software for logistics and supply chain companies.
- The system is mostly integration: carriers, marketplaces, customer portals, IoT devices, with ordinary accounting underneath.
- You have outgrown packaged options and can say precisely which of your processes they could not handle.
Even then, we rarely recommend writing your own general ledger. The most practical custom ERP for a small or mid-sized business is usually a hybrid: keep accounting in a proven package, and build the operational system around it, connected through the package's API. The same reasoning applies to customer management, which we cover in our look at when a custom CRM is worth building.
What each path costs over five years
Compare total cost over five years, not the first-year price. Here is the method using our distributor with 25 users. The numbers below are illustrative placeholders, not quotes or market prices; replace them with the real figures you are given.
Buy (subscription): 25 users × $40 per user per month × 60 months = $60,000. Add implementation, data migration and training, say $15,000. Five-year total: about $75,000, plus the harder-to-price cost of changing processes.
Customize (open-source base, self-hosted): no per-user license fee for the community editions. Implementation plus two custom modules, say $35,000. Hosting at $150 a month, $9,000. Upgrades and support at $6,000 a year, $30,000. Five-year total: about $74,000.
Build (custom operations system with packaged accounting): build, say $90,000. For maintenance and small changes, a planning figure of 15 to 20% of the build cost per year gives $67,500 to $90,000 over five years. Hosting, $9,000. Keep the accounting subscription for a few finance users, say $6,000. Five-year total: roughly $172,500 to $195,000.
On cost alone, building loses here, and it usually will. It wins only when the better fit is worth more than the difference. If custom dispatch rules save two full-time roles or prevent errors that cost you customers, put a yearly figure on that and add it to the comparison. If you cannot name that figure, buy or customize.
If you trade in India, check GST e-invoicing early
Indian businesses have one more filter. Under Notification 10/2023-Central Tax, e-invoicing has applied since 1 August 2023 to taxpayers whose aggregate turnover exceeds ₹5 crore. Whichever path you choose, confirm how invoices will be generated and reported, and test it with your accountant before go-live. For a custom build, this is one more reason to keep invoicing in a package that already handles it, or to budget properly for the integration.
Answer these before you speak to any vendor
- Which five processes cost the most when they fail, and who owns each one?
- How many people will log in, and how many only need reports?
- Which existing systems must connect: web store, marketplaces, accounting, couriers, banks?
- What data has to move across, and how clean is it? Product and customer records are usually messier than people think.
- Who on your side can spend real time on this for three to six months?
If several of these are hard to answer, a short paid discovery phase before any build or purchase decision is money well spent; our article on why a discovery phase saves money explains what it should produce.
Your next step is the one-page list of five processes. Write it this week, show it to the two or three people who run those processes day to day, and correct it with them. Every conversation with a vendor, including us, will go faster and be more concrete because of it.
Once the page exists, our ERP development team can read it and tell you which path we would take in your position, even when that means buying a package and configuring it rather than building.
