Services
Industry
Resources
Company
Summarize this blog post with:
● Shop floor ERP integration connects machines, operators and quality checks to the ERP so production, quality and finance work from the same data.
● The core problem it solves: the ERP knows what was supposed to happen, the shop floor knows what actually happened, and nobody reconciles the two until month-end.
● Four data flows matter most: work order status, material consumption, labour and machine time, and quality results.
● Standard architecture follows ISA-95 layers, with an MES or edge layer between PLCs and the ERP. Direct machine-to-ERP connections rarely survive contact with reality.
● Realistic gains: costing accuracy within days rather than weeks, faster scrap identification, and a WIP balance that ties to the physical floor.
● Start with one production line and four data points. Plant-wide rollouts that begin with full scope usually stall.
Walk any plant floor in Chakan, Sanand or Hosur the kind of manufacturing operation SaasWorx works with on NetSuite ERP and you will find two versions of the truth.
The ERP has a work order that says 5,000 units at a standard cost. The floor has a supervisor's notebook that says the line ran short two operators on Tuesday, one batch got reworked, and the changeover took ninety minutes instead of forty.
Both are accurate. Neither is complete. And the gap between them is where manufacturing margin quietly disappears.
Shop floor ERP integration is the two-way flow of data between production systems and the ERP. Work orders, routings and material requirements flow down to the floor. Completions, consumption, labour time, machine time and quality results flow back up. Finance draws actual costs from that returned data rather than from estimates.
The word that matters is two-way. Many plants have a one-way flow where the ERP issues orders and receives a completion quantity at the end. That is not integration. That is a printout with a return receipt.
Get these right and most of the value follows.
1. Work order status. Released, started, in progress, paused, complete. Real timestamps, not end-of-shift entries. Without this, WIP in the ERP is a guess.
2. Material consumption. What was actually issued and consumed against each order, including scrap and rework. Backflushing at standard hides yield problems for months.
3. Labour and machine time. Who worked on what, for how long, and which machine ran. This turns a standard cost into an actual cost.
4. Quality results. In-process inspection, final inspection, non-conformances and dispositions, linked to the batch or serial they belong to.
Most plants have some version of the first two. The third and fourth are where the real reporting gaps sit.
The instinct is to connect machines directly to the ERP. It rarely works, because ERPs are transactional systems built for business documents which is exactly why NetSuite integration and automation projects are scoped around an intermediate layer, not a direct feed, since a PLC sends signals every 200 milliseconds and no ERP is built to consume that.
The reference model is ISA-95, which describes manufacturing systems in layers:
● Level 0 to 1: Sensors, actuators, PLCs
● Level 2: SCADA and HMI, supervisory control
● Level 3: MES or manufacturing operations management
● Level 4: ERP and business planning
Level 3 exists precisely to translate between machine time and business time. An MES or a lighter edge layer aggregates high-frequency machine data into meaningful business events: order started, 500 units produced, downtime of 22 minutes with a reason code.
A growing number of plants use a unified namespace approach, where an MQTT broker holds a structured, current view of plant data that any system can subscribe to. It reduces the point-to-point integration sprawl that makes plant IT so brittle.
The practical rule: the ERP should receive events, not signals.
Quality is often treated as a separate system with its own database and its own reports. That separation is expensive.
When quality data lives apart from the ERP:
● Scrap and rework costs appear as unexplained variances at month-end.
● A hold on one batch does not automatically block shipment of that batch.
● Customer complaints cannot be traced back to a production run without manual work.
● Cost of poor quality is never calculated, so it is never managed.
When quality data lives inside the same data model:
● A failed inspection creates a non-conformance record linked to the work order, the batch and the material lot the same lot-level traceability covered in Batch and Serial Traceability Software in Manufacturing.
● The disposition, whether scrap, rework or use as is, posts the cost to the right account automatically.
● Inventory status changes to blocked without anyone remembering to do it.
● Cost of poor quality becomes a monthly number the plant head sees.
That last point is the one finance directors underestimate. Most plants cannot state their cost of poor quality. The ones that can typically find it is between 2 and 5 percent of revenue, and most of it is addressable.
This is where the integration pays for itself.
Standard costing works well as a planning tool and poorly as a management tool, because it tells you what things should cost the same blind spot behind Stockouts & Excess Inventory in Manufacturing: How Inefficiency Erodes Margins. Integrated shop floor data tells you what they did cost, at the level of the individual order.
With four data flows connected, finance can produce:
● Actual cost per work order, split into material, labour, machine and overhead
● Yield variance by product and line, visible weekly rather than monthly
● Rate and efficiency variances that trace to specific shifts and machines
● True WIP valuation that matches what is physically on the floor
● Cost of poor quality, broken out by defect category
The behavioural change matters more than the reports. When a plant manager sees yield variance on Wednesday for Monday's run, the conversation is about causes. When they see it on the 12th of the following month, the conversation is about explanations.
Consider a mid-sized auto components manufacturer running three lines with a mix of machined and assembled parts.
Before integration, month-end works like this. Production reports quantities from a shared spreadsheet. Stores reconciles material issues against a physical count. Quality maintains its own log of rejections. Finance builds a costing sheet from all three, finds a variance it cannot explain, books it to a general absorption account, and closes.
After integration, the same month-end works differently. Completions and consumption post as they happen. Downtime carries reason codes. Rejections create linked non-conformance records with automatic cost postings. The variance report is populated on day two, and each line has an owner who already knows what it says.
Nothing about the physical process changed. What changed is that the data now arrives in the same system, at the same time, in a form finance can use.
1. Pick one line. Not one plant. One line, ideally your highest-volume or highest-margin one.
2. Instrument four data points. Order start and stop, quantity produced, quantity rejected, and downtime with reason codes. Resist the urge to capture forty.
3. Fix the master data first. Bills of materials, routings and work centre rates must be accurate before anything else means anything, a step covered in more depth in ERP for Manufacturing Industry in India. This step is unglamorous and non-negotiable.
4. Run parallel for one month. Keep the manual process alongside the new one. Compare. Investigate every difference, because each one is either a data problem or a process problem worth knowing about.
5. Connect quality next. Inspection results, non-conformances and dispositions flowing into the same records.
6. Then extend. Second line, second plant, additional data points, predictive maintenance signals.
Plants that follow this sequence usually see credible actual costing within a quarter. Plants that start with a full-scope, all-plants programme typically spend eighteen months and deliver a dashboard nobody uses.
Infor MES, Siemens Opcenter, Rockwell FactoryTalk, AVEVA MES, Critical Manufacturing and SAP Digital Manufacturing serve the full MES layer. Tulip is widely adopted for no-code frontline apps and is used where GxP requirements apply. MachineMetrics suits plants whose immediate pain is machine downtime and OEE visibility, and it can act as a first step towards fuller integration.
Oracle NetSuite covers work orders, routings, WIP and quality within the same platform for discrete and process manufacturers, including the AI-powered automation layer that turns captured shop floor events into forecasts and anomaly alerts, which suits mid-market plants that do not need a full separate MES. SAP and Microsoft Dynamics 365 occupy similar ground at different scales.
OPC UA remains the standard for machine data. MQTT is the common transport for unified namespace architectures. Both are worth specifying in any vendor conversation, because a platform that requires proprietary connectors becomes expensive at the second plant.
● Bad master data. If routings are three years old, integrated actuals will be compared against fictional standards. Fix this first.
● Operator burden. If a data capture step adds two minutes per order, it will be skipped or gamed. Automate capture wherever the machine can report it.
● Too many metrics. Plants that begin with fifty KPIs end with none. Four to six is the right starting number.
● No owner on the floor. Integration projects owned only by IT or only by finance fail. A production supervisor needs to want the data.
● Ignoring the network. Older machines, shop floor Wi-Fi dead zones and unmanaged switches account for a surprising share of stalled projects.
What is the difference between MES and ERP in manufacturing?
The ERP plans and records the business transaction: what to make, when, at what cost, and how it appears in the accounts. The MES executes and tracks production on the floor in real time: which machine, which operator, which batch, what stopped and why. The ERP works in business time, the MES in machine time.
Can we connect machines directly to the ERP without an MES?
For simple, low-volume operations, sometimes. For most plants, the frequency and format of machine data make a middle layer necessary. Direct connections tend to work in a pilot and break at scale, because the ERP is not designed to consume high-frequency signals.
How long does shop floor ERP integration take?
A single line with four data points is typically eight to twelve weeks, including master data cleanup. A multi-line plant runs six to nine months. Multi-plant programmes should be sequenced as repeated single-plant rollouts rather than one large project.
What is the first metric to track?
Yield by work order, compared against the bill of materials. It is simple to calculate, immediately meaningful to both production and finance, and it exposes data quality problems fast.
Does this replace our quality management system?
Not necessarily. It links to it. What matters is that quality events, batch identity and inventory status share one source, so a hold in quality automatically blocks the material in the ERP.
The gap between the plan and the floor is not a technology problem at heart. It is a timing problem, made worse by systems that were never asked to talk to each other.
Close it and the arguments change. The weekly production review stops being a debate about whose numbers are right and starts being a discussion about what to fix.
SaasWorx works with manufacturers on ERP and shop floor integration, costing structure, and the reporting layer that connects production, quality and finance. If your month-end still starts with a variance nobody can explain, book a consultation to find out where your data gap sits.
