Build vs. Buy ERP: A Scored Decision Framework and Its Disqualifiers
Buy when your processes are standard and your seat count is small; build when two or three workflows are the thing that makes you money; run a hybrid — licence the core, build the differentiator — for most cases in between. Below roughly 25 seats a licensed suite is usually cheaper across three years, and that is the threshold at which this page argues against building. The rubric below is scored, not rhetorical.
Vivek Mishra — Founder & Lead Architect, ERPStack
Quick Answer & Verdict
Building beats buying when process fit, seat growth and integration count all point the same way and you have a named engineering owner. Buying wins when your finance processes are close to textbook, when you file statutory returns in many countries, or when you are below roughly 25 seats. Score the five axes on this page; a middling total means hybrid, not build.
3-Year TCO Comparison
| Cost line, 3-year horizon | Buy: Oracle NetSuite | Buy: SAP S/4HANA | Build: ERPStack custom |
|---|---|---|---|
| Year-1 implementation | $80,000 - $250,000+ | $150,000 - $500,000+ | $25,000 – $250,000+ one-time |
| Annual licensing (indicative band) | $25,000 - $150,000+ / year | $100,000 - $1,000,000+ / year | $0 — no seat licences at any headcount |
| Cost of 50 more users in year 2 | Priced by seat count; the band above was collected for mid-market seat deployments | Priced by deployment size; the band above reflects mid-market contract values | $0 — headcount is not a licensing event |
| 3-year total (year 1 + 3 years of licence) | $155,000 - $700,000+ | $450,000 - $3,500,000+ | $25,000 – $250,000+ plus hosting and maintenance |
| What you hold at the end of year 3 | A renewal quote | A renewal quote | Source code, the PostgreSQL database and the IP, under a perpetual licence granted at kickoff |
What the ERPStack column actually is: Next.js and React in TypeScript, PostgreSQL with Drizzle ORM, deployed to Vercel or AWS. No seat licences, no vendor API quota, and 100% of the source code, the PostgreSQL database and the IP transferred to the client under a perpetual licence granted at kickoff.
How to read the competitor columns: Where this table states an annual licence figure it is the same indicative band listed below and on that vendor's /alternatives page, with its basis and collection date stated there; any multi-year total is computed from it rather than quoted separately. Any implementation or customisation figure is ERPStack's own estimate for a mid-market deployment, not a vendor list price — several of the platforms listed below publish no list pricing at all.
Migration Timeline to Custom ERPStack
1. Score the five axes, then stop or continue
Score process non-standardness, 36-month seat growth, integration count, regulatory surface and in-house engineering capacity from 0 to 4 each, then add them. A total of 7 or below means buy the licence, and the honest end of the engagement is here rather than at a proposal.
2. Price both routes on one horizon
Put the licence quote you are holding beside the 3-year table above, then run your own seat count through the ROI calculator on this site. A build has to beat the licence stream across three years, not win on every line in year one.
3. Pilot the single module the score points at
Build the one workflow that scored highest while the incumbent stays the system of record. A single-module build is the cheapest way to falsify the score before the full scope, the data migration and the cutover are committed.
Hidden Platform Overheads to Avoid
The licence line compounds with headcount, not with usage
A licensed mid-market suite bills $37–$165 per user per month — ERPStack’s own observed list-price range across comparable mid-market SaaS suites, not a rate published by any single vendor. At 25 seats that is $11,100 to $49,500 a year, and three years of it is $33,300 to $148,500. Hiring is what moves that number, so the cost of the system is set by a decision the finance team does not make.
A custom build moves statutory maintenance onto your own backlog
SAP S/4HANA ships vendor-maintained statutory localisation packs and updates them as the law changes. That work does not disappear when you stop paying for it — it becomes a permanent line on your own roadmap. Budget it as recurring engineering rather than as a project cost, and price it before the build, not after.
The hybrid has a seam, and the seam has a cost
Licensing the core ledger and building the differentiator is the right answer more often than either pure route, and it is not free: you own an integration, a reconciliation job and a failure mode neither vendor supports. Price the seam explicitly. A nightly sync that silently drops rows costs more than either pure answer.
A build with no named owner is the most expensive option on this page
Source code you cannot staff is worse than software you rent, because the maintenance obligation arrives whether or not anyone is there to meet it. If no named person will own schema changes, releases and dependency upgrades after handover, the answer is buy — whatever the other four axes score.
Where the incumbent genuinely wins
A comparison that never concedes a point is marketing, not evaluation. These are the cases where buying beats building — read them before the verdict.
Speed to a compliant general ledger
Oracle NetSuite reaches a working, audit-familiar general ledger faster than any custom build cycle, and it arrives with an auditor population that already knows the system. If your constraint is a filing date rather than process fit, that gap settles the question on its own and no rubric total should be allowed to override it.
Vendor-maintained statutory localisation
SAP S/4HANA ships country-specific tax, payroll and statutory reporting packs and maintains them as the law changes. A custom ERP has to build each one and then keep maintaining it. Filing in many jurisdictions is the most reliable reason on this page to buy rather than build.
The hybrid beats both pure answers more often than either
Licence the ledger, build the differentiator. Re-deriving a general ledger you could rent is waste; renting the two or three workflows that actually make you money is worse. Most totals land in the middle band and resolve to a hybrid, and a page that concluded "build" at every score would be exactly as useful as the vendor pages that conclude "buy".
When a custom build is the wrong answer
- Seat growth: fewer than roughly 25 users and no compounding headcount plan. At the low end of observed per-seat pricing a licence stream that small does not repay a build across three years — the same threshold this site publishes on its pricing page.
- Process fit: a mid-market suite already runs your process without modification. Configuration beats code every time configuration is genuinely available.
- Regulatory surface: statutory filings across many countries. Vendor-maintained localisation is worth more than source-code ownership when the obligation changes without notice.
- Engineering capacity: no named internal owner for schema, releases and dependency upgrades after handover. This is a hard gate and it overrides every other axis.
- Integration count: one or two integrations, each with a vendor that publishes a supported connector for the suite you are already considering.
- Timeline: a hard statutory or contractual go-live inside two months for a full multi-module finance suite. Nothing in our published delivery window makes that safe.
Which Option Should You Choose?
Score the five axes and act on the total rather than on the preference you walked in with. Seven or below: buy the licence, and this page is telling you not to hire us. Eight to thirteen: hybrid — licence the core ledger, build the two or three workflows a template cannot express. Fourteen or above, with a named engineering owner already in place: build, and expect the licence stream you avoid to be the line that repays it. If the engineering-capacity axis scores zero the total is irrelevant and the answer is buy, because owning source code you cannot staff is the most expensive outcome on this page.
Annual licence benchmarks across every platform we compare
Indicative annual licence bands for the platforms this site maintains a full analysis of. They are ERPStack's own estimates — read off the vendor's pricing page where one exists, and off prospect-supplied quotes where it does not — so the basis and the collection date for each band are printed beneath it. Use them to sanity-check the table above against whatever quote you are holding.
- Oracle NetSuite$25,000 - $150,000+ / year
ERPStack estimate for 25-250 seats — not an Oracle quote and not a list price. No Oracle-published NetSuite figure was readable on 15 August 2026 (netsuite.com refuses automated requests), so the floor is anchored to the nearest suite that does publish one: Dynamics 365 Business Central at $80-$110 per user per month is about $24,000 a year at 25 seats (microsoft.com/en-us/dynamics-365/products/business-central/pricing, read 15 August 2026). Get the base licence, the modules, and the year-2 and year-3 renewal price in writing.
- SAP S/4HANA$100,000 - $1,000,000+ / year
ERPStack estimate for mid-market and upper-mid-market deployments — not an SAP quote. No SAP-published S/4HANA figure was readable on 15 August 2026 (sap.com refuses automated requests), so this is an order-of-magnitude estimate of annual licence plus maintenance, excluding implementation. What SAP does publish is the arithmetic beside it: extended maintenance for Business Suite 7 costs "a premium of two percentage points on the existing maintenance basis" (news.sap.com, read 15 August 2026). Get licence, maintenance and digital-access documents quoted separately, in writing.
- Odoo Enterprise$5,000 - $30,000+ / year
Derived from Odoo published per-user Enterprise pricing plus typical implementation-partner fees, at 25-250 seats. Verify current per-user pricing at odoo.com/pricing. Collected July 2026.
- Salesforce CRM & Commerce$50,000 - $500,000+ / year
Derived from Salesforce published Sales Cloud edition pricing (Professional / Enterprise / Unlimited) at 25-250 seats, plus typical add-on and integration licensing. Verify at salesforce.com/pricing. Collected July 2026.
- Microsoft Dynamics 365$20,000 - $180,000+ / year
Derived from Microsoft published Dynamics 365 per-user licensing at 25-250 seats, plus typical partner implementation fees. Verify at microsoft.com/dynamics-365/pricing. Collected July 2026.
- Epicor ERP$30,000 - $200,000+ / year
ERPStack estimate for mid-market manufacturing at 25-250 seats — not an Epicor quote. Epicor publishes no Kinetic price on its own product page: the calls to action there are "Request a Demo" and "Take a Product Tour" (epicor.com/en-us/erp-systems/kinetic, read 15 August 2026). Treat the band as an order-of-magnitude estimate of annual licence plus maintenance, and get the licence, the maintenance uplift and the partner implementation fee quoted separately, in writing.
- Sage X3$15,000 - $100,000+ / year
ERPStack estimate for mid-market deployments at 25-250 seats — not a Sage quote and not a list price. No Sage-published X3 figure was readable on 15 August 2026: sage.com refuses automated requests, which is why no Sage price is linked here. Treat the band as an order-of-magnitude estimate of annual licence plus maintenance, and get the licence, the renewal uplift and the partner implementation fee quoted separately, in writing.
- Zoho Creator$3,000 - $25,000+ / year
Derived from Zoho published Creator per-user pricing at 25-250 seats, plus typical implementation effort. Verify at zoho.com/creator/pricing. Collected July 2026.
- Monday.com Work OS$8,000 - $60,000+ / year
Derived from monday.com published per-seat tier pricing at 25-250 seats, plus typical enterprise add-ons. Verify at monday.com/pricing. Collected July 2026.
- HubSpot CRM & Ops Hub$15,000 - $120,000+ / year
Derived from HubSpot published Sales Hub and Enterprise tier pricing at 25-250 seats, plus typical onboarding fees. Verify at hubspot.com/pricing. Collected July 2026.
Frequently Asked Questions
How do we score build vs buy without guessing?
Score five axes from 0 to 4 and add them. Process non-standardness: 0 if a template fits, 4 if the workflow is the product. Seat growth: 0 below roughly 25 users over 36 months, 4 if headcount compounds. Integration count: 0 for one or two supported connectors, 4 for many bespoke systems. Regulatory surface: 0 for statutory filings across many countries, 4 for one jurisdiction with unusual internal controls. In-house engineering capacity: 0 with no named owner, 4 with a team that already ships. Seven or below, buy. Eight to thirteen, hybrid. Fourteen or above, build.
What disqualifies a custom build whatever the total says?
Two hard gates. If in-house engineering capacity scores zero — no named person owning schema changes, releases and dependency upgrades after handover — buy, because source code you cannot staff is worse than software you rent. If your seat count is below roughly 25 and not compounding, buy: three years of licences at 25 seats is $33,300 to $148,500 — ERPStack’s own observed list-price range across comparable mid-market SaaS suites, not a rate published by any single vendor — against a build starting at $25,000 before hosting and maintenance. At the bottom of that range the avoided licences barely clear the bare build and fall short once hosting and maintenance are added, which is why the threshold is published as roughly 25 seats rather than as an exact crossover.
When is the hybrid right rather than build or buy?
Whenever the total lands between eight and thirteen, which is most mid-market operations. Licence the general ledger, accounts payable and payroll — the places where being different is a liability — and build the two or three workflows a template cannot express. What you accept in exchange is an integration seam: a sync job, a reconciliation and a failure mode neither vendor supports. Price that seam before you commit, because it is the line that decides whether the hybrid actually beats the pure route it replaced.
Does a custom build ever cost less than the licence in year one?
It can, depending on the implementation fee attached to the licence, and it is the wrong question. Year one loads the build and lightly loads the licence, so the comparison only becomes meaningful across three years — which is how the table on this page is drawn: implementation once, licences three times. What decides it is whether your licence line compounds with headcount. If it does, a crossover arrives; if headcount is flat and small, it never does. Run your own seat count through the ROI calculator on this site rather than trusting either column here.