Custom ERP vs Odoo: how to actually choose for a growing SMB

Summary: Choose Odoo / ERPNext off-the-shelf for standard processes, extend open source for a few differentiating workflows, or a custom ERP when the process is the product—then roll out in phases so users actually adopt.

"Should we buy an ERP, customize an open-source one, or build our own?" is the question every operations director asks at some point — usually around the time the spreadsheet count crosses 30 and the warehouse starts disagreeing with accounting.

The honest decision tree

• Use Odoo / ERPNext as-is if more than 80% of your processes are textbook (standard sales, simple stock, generic accounting) and your industry is well-supported.
• Extend an open-source ERP if you have 2–3 processes that are genuinely your competitive edge (custom pricing, niche logistics, specific compliance) and the rest is standard.
• Build a custom ERP if your business is the process — marketplaces, multi-tenant operations, regulated industries, or anything where the ERP is the product.

A phased digitalization roadmap that actually ships

The biggest ERP failure mode is a 12-month project that goes live as a big bang and nobody uses. The fix is boring:

1. Weeks 1–3: map current processes with the people doing them. Find the top 3 sources of pain — almost always quoting, stock, or billing.
2. Weeks 4–10: ship one module that fixes one of those pains end-to-end. Real users, real data, weekly demos.
3. Months 3–6: add adjacent modules (accounting integration, inventory, CRM) once the first one is trusted.
4. Months 6+: connect to ecommerce, BI dashboards, and the AI assistants that turn your ERP data into decisions.

Signals you actually need a custom ERP

• You explain your business by saying "it's like X, but...", and the "but" is more than 20% of revenue.
• Your team maintains parallel systems because no off-the-shelf ERP fits — that duplication is your real cost.
• You want the ERP, the ecommerce, the customer portal and the AI assistants to share one source of truth, with no daily CSV exports between them.

If two or more of those apply, custom (or heavily-extended open source) usually wins on a 3-year horizon, even though it looks more expensive on day one.