Your ERP Is a Filing Cabinet, Not a Business Brain

Your ERP knows what happened.

That does not mean it knows what your business should do next.

I have walked into companies with a major ERP, years of historical data, and dashboards covering every department. Leadership believed the operation was connected because everything important was supposed to be in one system.

Then we followed a real decision from beginning to end.

The ERP held the transactions. The forecast lived in a spreadsheet. The production plan depended on a file maintained by one experienced employee. A supplier delay was visible on a dashboard, but the response still had to be worked out manually across purchasing, production, labor, and customer commitments.

The company had a system of record. It did not have a complete operating picture.

Here is the simplest way I can put it: an ERP is a digital filing cabinet. It can be essential infrastructure without being the brain of the business.

01 What an ERP is actually designed to do

An enterprise resource planning system is built to integrate and record core business activity. Orders, invoices, inventory movements, financial entries, purchasing records, employee data, and production transactions all need a controlled home. That is what makes the ERP foundational.

The mistake is not buying an ERP. The mistake is assuming that recording the operation is the same as understanding it.

A transaction tells you that material arrived late. It does not automatically know every consequence of that delay. It may not know which machine can take a different job, which trained operator is available, which customer commitment has the highest penalty, or which production sequence protects the most margin.

Those decisions require context. Context usually spans systems, departments, working files, and human experience.

This is one of the most important ERP system limitations: the software can contain accurate records while still holding an incomplete model of how the company really operates.

02 The spreadsheet beside the ERP is not an exception

Almost every leadership team tells us some version of the same story: “The ERP is our source of truth, but this department still needs a spreadsheet.”

That “but” matters.

Researchers have documented the pattern for years. A 2018 study of production planning and control found that spreadsheets continued to dominate the daily work of planners even where ERP and advanced planning systems were available. The authors examined three cases and concluded that spreadsheets remained the main practical support for capacity planning and production scheduling (IFAC-PapersOnLine, 2018).

A newer 2025 multiple-case study across eight manufacturers found the same basic pattern. ERP systems acted as the planning backbone, but teams repeatedly supplemented them with Excel for flexibility, especially in engineer-to-order and assemble-to-order environments. That flexibility helped planners react, but it also introduced data-consistency and scalability risks (IFAC-PapersOnLine, 2025).

Another mixed-method study found a substantially poorer fit between ERP functionality and decision-support requirements in make-to-order companies than in make-to-stock environments (Computers in Industry, 2015). When demand, routing, lead times, and production sequences change frequently, a fixed workflow struggles to represent every decision the operation must make.

So employees build the missing layer themselves.

They create trackers, calculators, macros, local applications, and unofficial handoffs. These tools keep work moving, but create a second operating reality beside the ERP.

03 Shadow systems are evidence, not just bad behavior

The research calls these unofficial tools shadow systems. They are applications, spreadsheets, databases, or services used by business teams outside formal IT management.

In a multiple-case study of established ERP environments, researchers found that 64% of the shadow systems they examined were partially or fully dependent on ERP data or functionality. In other words, the unofficial tools were not doing unrelated work. They were extending, duplicating, or compensating for parts of the official enterprise system (Systems, 2016).

If a critical planning file exists outside the ERP, the first question should not be, “How do we force everyone back into the system?”

The better questions are:

• What decision does this file make possible?
• What context is missing from the ERP?
• Which rules or assumptions live only inside this file?
• Who understands how it works?
• What happens if that person is unavailable?

Workarounds often solve an immediate operational problem. Research on ERP workarounds describes how they can improve resilience in the short term, but become harmful when they hide structural weaknesses and prevent the organization from correcting them (Continuity & Resilience Review, 2020).

The spreadsheet that saves today’s production plan may also be the reason nobody fixes the underlying information gap.

04 Your dashboard may still be a rear-view mirror

This is where many teams push back.

“We already solved that. We have business intelligence dashboards on top of the ERP.”

Dashboards are valuable. But a dashboard does not automatically create intelligence simply because it visualizes data.

IBM describes four levels of analytics:

• Descriptive: What happened?
• Diagnostic: Why did it happen?
• Predictive: What might happen next?
• Prescriptive: What should we do next?

Prescriptive analytics is different because it recommends a course of action while considering constraints, objectives, uncertainty, and tradeoffs (IBM, “What Is Prescriptive Analytics?”).

Most operational dashboards remain concentrated in the first level. They show yesterday’s output, this month’s costs, current inventory, or delayed orders. Some allow users to drill into causes. Far fewer model the future and recommend a coordinated response.

That is why I describe many dashboards as rear-view mirrors. A rear-view mirror is useful. You should not remove it. But you would never steer the company using only a view of the road behind you.

There is another limitation. A dashboard can only be as complete and current as the data feeding it. If production data is disconnected from workforce availability, supplier status, quality events, or a planner’s working file, the dashboard can look precise while omitting the constraint that changes the decision.

05 What happens when one material is late

Take a production planner managing a factory schedule.

At a minimum, the planner is balancing three categories of information:

• Machines: Which assets are available, capable, and properly tooled?
• People: Which qualified operators are available for the required process?
• Materials: Are the inputs available in the right quantity at the right time?

Now one supplier misses a delivery.

The ERP may record the revised date. The dashboard may flag the affected material. But the real decision is much larger:

• Which orders can still run?
• Which jobs should move to another machine?
• Is the right operator available for the new sequence?
• Which setup changes create additional downtime?
• Which customer commitment carries the greatest risk?
• What happens to every downstream production step?

NIST research on dynamic manufacturing systems makes the same requirement explicit. Responding to maintenance issues, workforce changes, or upstream delays requires interoperability between shop-floor systems and higher-level systems such as MES and ERP. Those systems need visibility into the state of relevant assets, and the exchanged information must contain enough production context to support decisions (NIST, Current State and Future Roadmap of Distributed Manufacturing Systems, 2020).

The same report identifies three persistent barriers: system complexity and disconnectedness, a lack of interoperable models, and a lack of contextual interoperability.

That last phrase is the important one.

Moving data is not enough. The receiving system has to understand what the data means in the operation.

06 The key-person risk hiding inside the process

When systems cannot model the full decision, an experienced person closes the gap.

That person knows which supplier is usually late, which machine can handle an unofficial substitution, which operator can solve a difficult setup, and which customer will accept a partial shipment. Years of operational memory become the real decision engine.

This expertise is valuable, but fragile when it is undocumented. The goal is not to replace experienced people with software. It is to capture their rules, constraints, exceptions, and escalation paths so the organization can use them consistently.

If a planning process stops when one employee goes on leave, that is not only a staffing problem. It is an architecture problem.

07 What a system of intelligence adds

A system of intelligence sits above the systems of record. It connects operational signals, adds business context, and helps the organization move from reporting to decision support.

It should be able to combine:

• Trusted transactions from the ERP
• Production status from MES and shop-floor systems
• Machine and sensor signals
• Procurement and supplier information
• Workforce availability and qualifications
• Rules currently embedded in spreadsheets
• Exceptions and domain knowledge held by experienced employees

For the late-material example, the intelligence layer should not stop at “delivery delayed.” It should evaluate feasible production sequences, identify affected orders, model the tradeoffs, and recommend the strongest response. Where the decision is repeatable and governed, a system of action can then execute approved steps back through existing tools.

The ERP remains the official record. It does not need to be replaced. It needs to become part of a larger decision architecture.

08 AI makes the missing context more dangerous

An AI system cannot reliably reason over context it cannot access. If the ERP contains only part of the operation, the model sees only part of it. If departments define the same measure differently, the model inherits that disagreement. If the latest plan lives in a private spreadsheet, the model works from outdated reality.

AI does not turn fragmented information into truth by default. It can produce a faster, more confident answer from the same incomplete foundation.

AI readiness should therefore begin with operational diagnosis, not tool selection.

09 Seven questions to ask before adding another dashboard

Start with one real, recurring decision and trace it.

• Which systems contribute information to the decision?
• Which spreadsheets or local tools are involved?
• How much information is re-entered or reconciled manually?
• How quickly does the ERP reflect a real-world change?
• Which rules depend on an employee’s memory or judgment?
• Does the dashboard explain what happened, or recommend what to do?
• Can the selected action flow back into the systems where work happens?

Do not begin by asking how to move every spreadsheet into the ERP. Begin by asking why each spreadsheet exists. Some should disappear. Some should be integrated. Some contain valuable operating logic that needs to be preserved before the file itself is retired.

10 Your ERP is essential. It is still not enough.

This is not an argument against ERP systems.

An ERP remains one of the most important platforms in the enterprise. It provides control, consistency, and a trusted record of transactions. The problem begins when leadership treats that record as a complete representation of the business.

Research and factory-floor reality point in the same direction. Planners still rely on spreadsheets, shadow systems remain tied to ERP processes, and descriptive dashboards stop short of recommending the next move. The answer is a connected data foundation, an intelligence layer that understands operating context, and a governed path from insight to action.

Your ERP can tell you where the business has been.

The next competitive advantage is building a system that helps you decide where it should go.

11 Frequently asked questions

What are the main limitations of an ERP system?

An ERP is optimized to integrate and record core transactions. Its limitations appear when a business expects it to capture every local workflow, incorporate real-time context from every operational system, model uncertainty, or recommend a coordinated response. Those needs often require connected planning, analytics, and intelligence capabilities around the ERP.

Why do companies still use spreadsheets after implementing an ERP?

Spreadsheets provide speed and flexibility when the ERP does not match a local planning requirement. Peer-reviewed manufacturing studies have found planners using spreadsheets alongside ERP and advanced planning systems for capacity planning, scheduling, and high-variability production. The spreadsheet is often compensating for a fit or context gap.

Does business intelligence solve ERP limitations?

It solves some of them. BI makes data easier to explore and communicate. However, many dashboards remain descriptive, answering what happened. Diagnostic, predictive, and prescriptive analytics require broader data, stronger context, and models that can explain causes, forecast outcomes, and recommend actions.

What is the difference between a system of record and a system of intelligence?

A system of record stores trusted transactions and historical events. A system of intelligence combines that record with additional operational data and business context to explain conditions, evaluate scenarios, and recommend what to do next.

How do we know if our ERP is missing critical operational context?

Follow a high-value decision through the company. If the decision requires manual exports, private spreadsheets, repeated reconciliation, or knowledge available from only one employee, the ERP is not carrying the full context required to make that decision.

12 Find the blind spots your ERP cannot see

At Embiggen, we start by mapping how decisions are actually made across the operation. We identify which data is trusted, which systems are disconnected, where spreadsheets carry hidden business logic, and where manual intervention is costing time or margin.

Then we connect what already exists and build the intelligence layer around it.

Book a free assessment and find out what your ERP is not showing you.

13 Sources

• Huber, M., Zimmermann, S., Rentrop, C., and Felden, C. “The Relation of Shadow Systems and ERP Systems: Insights from a Multiple-Case Study.” Systems 4, no. 1 (2016). doi.org/10.3390/systems4010011
• Jonsson, P., et al. “Spreadsheet Application Still Dominates Enterprise Resource Planning and Advanced Planning Systems.” IFAC-PapersOnLine 51, no. 11 (2018), 1224–1229. doi.org/10.1016/j.ifacol.2018.08.423
• “Exploring Tactical Production Planning Practices: Master Data Quality, Update Practices, and Planning Tools.” IFAC-PapersOnLine 59, no. 10 (2025), 2082–2087. doi.org/10.1016/j.ifacol.2025.09.350
• Deep, A., Guttridge, P., Dani, S., and Burns, N. “The Applicability and Impact of Enterprise Resource Planning Systems: Results from a Mixed Method Study on Make-to-Order Companies.” Computers in Industry 70 (2015), 127–143. doi.org/10.1016/j.compind.2014.10.003
• “Analysis of Enterprise Resource Planning System Workarounds with a Resilience Perspective.” Continuity & Resilience Review 2, no. 2 (2020), 131–148. doi.org/10.1108/CRR-06-2020-0022
• Helu, M., Morris, K., Jung, K., Lyons, K., and Leong, S. Current State and Future Roadmap of Distributed Manufacturing Systems. National Institute of Standards and Technology, 2020. nist.gov publication PDF
• IBM. “What Is Prescriptive Analytics?” ibm.com/think/topics/prescriptive-analytics