Building ERP from Scratch
retained text
Not looking to sell anything. I run a 15 entity business with holdcos and topcos. Currently using Xero Healthcare, software, real estate verticals mainly Bottom end of market: quickbooks, sage, Xero - can’t do multi-entity Top end of the market: net suite, dynamics 365, SAP - just clunky, will only use 10% if its features, and goodness the pricing Middle of the market: sage intact, SAP business one, AI native (campfire, rittle, dualentry) - still too expensive Only fit was Odoo enterprise But I really liked the UI and feel of the AI native platforms So I’ve started: \- ERDs (data diagrams mapping each table and field and its dependencies) and workflows are key. Built a 20 table ERD. We don’t need many things like inventory valuation, our asset list is very simple (maybe 400-500 items) \- Mapped and migrated COAs, Invoices, Bills, Contacts, 70k line items from general ledger detail and opening TB balances \- Mapped security controls, lock-in periods, inter companies Front end - Vercel Hosting - Cloudflare Backend - Supabase Iterations with Claude. I have 2 full time devs. An experienced FC who has done multiple ERP migrations for billion dollar companies. Reason I mention this is I saw overwhelming advice to not build your own ERP. Technical debt Maintenance debt Staying up to date with new tax/ policy changes Technical debt I’ve become very involved in the software and finance side of the business in the last 2-3 years. I look after both functions and built every workflow and system ground up so feel very comfortable with the granular finance mappings. Maintenance debt 2 full time devs. 40% of their workload is debugging. System issues in finance is unacceptable. But they are there for that reason. We have continuity fail safes, second by second data back-ups and good backup architecture and version restoration. Staying up to date Our accountants still do our year end Tax compute is off the books And any incremental changes, I plan to route in through fractional advisory or looping in with key data repositories We have a working prototype Plaid connector for bank recs Migrations pattern mapped for rollout Forgive my naivety, but for our business which is relatively simple billing patterns, minimal asset list - it doesn’t seem hard at all. The only complexity is inter-company which is c. 20 transactions a month and provisions for allocating invoices/ bills over multiple periods. I was worried seeing the overwhelming sentiment not to take this path. But the deeper we go into it, the more we are convinced ironically. Because, we won’t commercialise it - I have obviously stolen a lot of ideas from other sandboxes from other ERPs. ERDs will be a straight transfer from Xero so no field changes which will save a lot of headache. Sure we are uniquely positioned, and I don’t recommend anyone do this without proper resources. But at the same time, for those who are really in the detail with lean finance teams, with developer resource - I don’t know why many people aren’t going down this path. I’ve in 3 months built an all-in-one odoo type healthcare platform which works better than almost anything on the market. It’s commercialised - a lot of debugging on go live but it is nearly, if not, more complex than an ERP due to it having every function of the business on the application. My silly brain tells me - it’s just a bunch of ERDs, workflows, security controls, RLS, 10s of thousands of rows of data which map cleanly to the direct crossover of fields (no field to field mapping). Honestly not as complicated relatively to other platform builds. Idk. Anyway, I’ll let you know how it goes but really up to hear any feedback or suggestions from anyone.
document fetched 2026-10-08 00:18 · canonical https://reddit.com/r/Accounting/comments/1x0311m/building_erp_from_scratch