A full ERP implementation for a $25M discrete manufacturer in 2026 lands between $450,000 and $900,000 over twelve to eighteen months. Consulting and customization absorb the majority of that spend.
Controllers revisit the number when they try to square a WIP variance against a work order whose routing drifted after the last plant layout change.
The budget below is sized for a company with one headquarters, two production plants, 60 to 90 named users, shop-floor data collection, work-in-process tracking across routings, and multi-plant inventory netting. It is the profile most controllers mean when they say "mid-market manufacturer," and it is where the old ERP pricing formulas start to misprice the project.
Where the Headline Number Actually Comes From
The median ERP project across industries runs about nine months and $450,000 in direct cost, according to the 2025 Panorama Consulting ERP Report. That median is useful as a floor, not a planning figure, because it blends professional services firms and small distributors into the same mean as discrete manufacturers carrying WIP. For a manufacturer, the planning figure sits above it.
Software subscription itself is the smallest moving part. SAP S/4HANA Public Cloud is published at $180 per user per month, which annualises to about $150,000 for 70 users. That is the line the CFO sees first and the one that drifts least. Everything above it, services, integration, customization, data migration, change management, is where the project actually succeeds or blows up.
What a $25M Discrete Manufacturer Actually Pays For
A defensible 2026 budget for this profile breaks down roughly as follows. Software licensing sits at $130,000 to $180,000 in year one, depending on tier and user count. Implementation services, the systems integrator or vendor professional services team, land at $180,000 to $360,000 at a typical multiplier of one-and-a-half to two-and-a-half times first-year subscription.
Customization adds another $70,000 to $160,000 when the manufacturer runs anything unusual: a custom sales configurator, a non-standard quality hold flow, or a costing method the base system does not support cleanly. Shop-floor integration, meaning labor collection terminals, machine tag reads, and barcode scanners talking to the ledger, runs $40,000 to $90,000 per plant. Data migration, mostly the chart of accounts rebuild and the open work order cutover, sits at $30,000 to $70,000. Training and change management, if funded properly rather than skimped, should land near $60,000 to $100,000.

Shop Floor Integration Is the Line That Breaks Budgets
The shop-floor line is where mid-market manufacturing projects most often slip. A $25M manufacturer with two plants typically has somewhere between twelve and forty data collection points: handheld scanners at receiving, tablet terminals at each work center, and either OPC-UA or Modbus taps on the CNC or injection machines that matter for throughput. Each one is a small integration with its own clock-sync and error-handling conversation.
Prosci's analysis of why ERP implementations fail puts weak scope control and under-funded change management ahead of software selection. On the manufacturing floor that translates directly: the routing a process engineer draws on a whiteboard is not the routing the supervisor actually runs, and the mismatch only surfaces during user acceptance testing on live work orders. Budgets that assumed a clean mapping absorb the delta as change requests at $180 to $240 per billable consulting hour.
WIP Costing and Multi-Plant Inventory Are Their Own Line Item
WIP costing is the second predictable blow-up. A discrete manufacturer carries raw material, in-process assemblies at partial labor burden, and finished goods, all moving between plants under inter-company transfer rules. If the new ERP expects standard cost and the old one ran actual, or vice versa, somebody rebuilds the costing bridge. That rebuild is quiet work that lands squarely on the controller. Writing a chart of accounts migration plan that preserves variance analysis across the cutover protects three to six weeks of month-end sanity on the other side.
Multi-plant inventory netting carries similar tax. A standard SKU in two plants with different min-max levels and a transfer order rule between them is three master records, two policies, and one integration contract, not one SKU. That multiplication is where customization creep begins. The deterministic ledger inside erp.io models the master once and lets agents handle the plant-level policy variation without a schema change for each one.
Where Role-Based Agents Compress the Consulting and Customization Lines
The consulting and customization lines are where agentic ERPs make the budget move. Traditionally, a systems integrator writes a specification, a developer builds a custom field or workflow, QA tests it, and the controller signs off. That cycle is billed at a blended rate and repeats for every exception the business cares about.
With role-based agents absorbing AP coding, AR matching, inventory reconciliation and plant-level cost variance review, a meaningful share of what would have been custom code becomes agent policy instead. The AP agent reads the invoice, codes it against the purchase order, checks the three-way match, and posts, with the ledger logging the deterministic entries and the agent explaining its reasoning in the audit trail. That is work an SI would otherwise quote as a custom approval workflow. Published guidance on AP agent approval thresholds under SOX shows how the policy sits in configuration rather than code.
| Metric | Cloud advantage | |
|---|---|---|
| Payback period | Payback period | 43% faster |
| Return on investment | Return on investment | 42% higher |
| Initial personnel costs | Initial personnel costs | 50% lower |
| Consulting expenses | Consulting expenses | 80% lower |
The pattern replicates for the three other common line items. Shop-floor exception review, where a supervisor currently calls a planner about a short labor ticket, moves into agent-mediated resolution with the controller reviewing exception summaries rather than individual transactions. Multi-plant inventory netting runs on agent-executed rules that the finance team adjusts, not SI-coded workflows. Daily reconciliation, instead of a month-end scramble, runs as the AR agent closes each day. The compound effect, which Nucleus has quantified in the context of cloud versus on-premise across ROI, payback and consulting cost, is a structural reduction in the consulting envelope rather than a discount.
A 2027 Budget That Holds Up in Board Review
The number that goes into the board memo should be a range tied to scope assumptions the board can read. For a $25M discrete manufacturer planning a 2027 go-live, a defensible top line is $450,000 to $900,000, with two plants, shop-floor integration at both, WIP costing migration, and change management funded as a real line. The lower bound assumes an agentic platform absorbing AP, AR and reconciliation work; the upper bound assumes a traditional SI-led project with heavier customization.
Two scope conditions decide which end of the range the project lands on. The first is how many custom fields and workflows the business insists on carrying forward from the legacy system, which is the lever that turns custom fields into ERP debt faster than most teams expect. The second is whether the integration contract between the ledger and the plant floor is treated as a vendor deliverable or as a jointly owned artifact, since migrations run with reconciliation proof catch divergence on day one rather than during month four of UAT.
Published customization-risk analysis in the International Journal of Computer Applications notes that customization defects cluster around fields that bridge operational and financial data, which is precisely where a discrete manufacturer's requirements pile up. That matches the field: the routings, the costing method, and the inventory netting rules are where the SI hours go, and where an agentic approach reshapes the budget.
The controller who started squaring a WIP variance at the top of this piece ends the implementation with the same question answerable from the dashboard the AR agent writes to each morning. The variance is traced, the work order is reconciled, and the budget presented to the board matches what was spent, because the consulting lines that used to swell the total are now agent policy the finance team controls directly.
