You don't need a new ERP. You need to see your floor.

Most ERP replacement projects start with a feeling, not a finding. Orders are late. Nobody can say where a job is without walking out to find it. The ERP gets the blame, and someone books a vendor demo. Before you sign anything, ask a narrower question: what exactly is the ERP failing to do? In my shops the answer was almost never accounting, purchasing, or inventory records. It was execution: what is running right now, what is waiting, and who needs to know. That is a different layer, and the people who write the standards for these systems have said so since the 1990s.^[1]

What your ERP is built to do

The ISA-95 standard, the international reference for connecting business systems to manufacturing, puts ERP at Level 4, "business planning and logistics," which it describes as "all the activities related to running a business."^[1]

That is the right job for it. Customer orders, bills of material, routings, standard costs, purchase orders, MRP, invoicing, the general ledger. These are records. They change in hours and days, not minutes. An ERP is a filing system with math attached, and a good one is worth keeping.

The Pennsylvania MEP center MANTEC makes the case plainly: a properly implemented ERP "functions as a single source of truth," ties business processes together, and lets you do job costing from recorded labor, material, and overhead instead of estimates.^[3] If your ERP cannot hold a routing or a work center, you have an ERP problem, and the last section covers that case. But most of the pain I hear about is not a records problem.

What the floor actually needs

Here is a typical morning in a shop that "needs a new ERP." Purdue MEP described an Indiana plant where "every morning, a supervisor would physically count parts in FIFO lanes," ERP visibility was limited, and production control meant highlighting spreadsheet cells pink to mark jobs as released.^[2] The same article says that for many small and mid-size manufacturers scheduling "is usually solved with a pencil and paper, or a complex spreadsheet."^[2]

MANTEC's ERP guide says the same thing from the other direction: many small manufacturers "still lack the ability to automatically capture shop floor data," so "every morning they spend time sorting out what has been done on previous shifts."^[3]

Notice what those shops are missing. Not a record of the order. They have that. They are missing answers to questions like: which jobs are at which work center right now, what is on hold and why, what did second shift leave half done, and who was told when the hot job stalled. Those are execution questions. They change by the minute. IMEC, the Illinois MEP center, puts it this way: manufacturers "rarely struggle because they have no data," they struggle because the information "is spread across accounting software, ERP, CRM, quality systems, shop-floor tools, and spreadsheets," and "the problem appears when a business question crosses departmental or system boundaries."^[4] A Missouri MEP report recorded a manufacturer's own words: "we collect tons of data ... how do we make good decisions with the data we collect (this is a huge problem)."^[5]

That matches my thirteen years of fixing late orders. Every building had the same pattern: good people, scattered data, and nobody looking at the same numbers.

The layer between the office and the floor

The standards community solved this on paper a long time ago. ISA-95 splits a manufacturing business into five levels.^[1]

ISA describes Level 3 as "manufacturing execution systems (MES) and other systems ... that manage manufacturing operations," and says the standard "primarily deals with the interface between levels 3 and 4."^[1] Part 2 of the standard exists specifically to define what passes between "level 3 manufacturing systems and level 4 business systems."^[1] The time-horizon column is mine, but it follows the standard's logic: the ISA-95 committee adopted levels "where key operations and applications interoperate in common time horizons and with common purpose," meant to "define logical integration boundaries."^[6]

The questions that eat your supervisors' first hour all live on row 3. A new ERP is a bigger, newer row 4. It does not move down a row because you paid more for it.

Where MES came from

The concept was established by the analyst firm AMR Research in 1992 "as the execution layer of manufacturing applications, which exists between the enterprise and control system."^[7] MESA International followed in 1997 with the definition that still gets quoted: "Manufacturing Execution Systems deliver information that enables the optimization of production activities from order launch to finished goods."^[7] MESA's early white papers defined eleven MES functions and diagrammed "the flow of information between Planning systems (MRPII / ERP) and MES."^[8] ISA-95 later described the same ground as eight common activities, among them detailed scheduling, dispatching, execution management, data collection, and tracking.^[7]

An industrial engineering dissertation that reviewed the early MES models summed up why the layer was needed: "Manufacturing execution is absent in traditional MPC models," and newer models "suggest an execution layer to be inserted between the planning and control layers."^[9] That gap is what shops feel when they blame the ERP.

One note: MESA International ceased operations effective 30 June 2026 and ISA has taken on its content, so old MESA links may fail.^[10] The model did not go away.

What a new ERP will and will not fix

Swapping the ERP is the most disruptive thing most shops can do to their systems. Here is what the record does and does not say.

The documented record

The clearest public evidence comes from government audits, because the auditors publish. In 2009 the U.S. Government Accountability Office reported that the Navy's ERP program, with a first increment estimated at about $2.4 billion, had "experienced about $570 million in cost overruns and a 2-year schedule delay."^[11] A year later GAO reviewed nine Department of Defense ERP efforts and found that "none of the four programs we assessed had developed a fully integrated master schedule."^[12][13] One of the four, the Air Force's Expeditionary Combat Support System, was cancelled in 2012, with the lack of an integrated schedule cited as a major cause.^[13]

Those are not 40-person shops, but the failure mode is familiar at any size: no real schedule, no definition of done, and a go-live date that drifts.

On the academic side, a 2023 systematic review of 55 peer-reviewed studies published between 2000 and 2022 concluded that "the failure rate of ERP systems is relatively high" and that "the success of these systems remains debatable."^[14] Its top five failure factors: lack of top management support, inadequate education and training, a mismatch between system and business strategy, lack of project management performers, and users unwilling to use the system.^[14] None of the five is "picked the wrong vendor."

The numbers you will hear in a sales call

You will also hear that "70 percent of ERP projects fail," or 50 percent, or 75 percent. I went looking for the primary source behind those figures and could not open one. The peer-reviewed review above gives no headline percentage. A university dissertation I found repeats a range of "40% to 90%" without tracing it to a study.^[15] The honest statement: ERP failure is common enough that auditors and researchers study it, the definitions of "failure" vary (abandoned, late, over budget, benefits never realized), and the percentages repeated online are hard to trace. Do not let a vendor scare you with one, and do not use one yourself.

What the MEP centers say about why projects stall

MEP advisers' warnings are more useful than any percentage. MANTEC reports that ERP implementations "can run six to nine months" on average and "a full year or longer" when management support fades, that "many ERP implementations have failed because of unrealistic timelines and expectations," and that a common failure is "running dual systems," which asks staff to do twice the work.^[3]

In a 2026 piece on why shop-floor technology projects fail, the same center lists four patterns: the system measures something nobody owns, data entry lands on the person with the least time, nobody defined finished, and the pilot ran on the easy line.^[16] Its advice is to keep "the existing management routine running in parallel" until the new system earns the handoff.^[16] Every one of those is a management problem, and every one will follow you into a new ERP.

The execution-layer approach

Here is the alternative. Keep the ERP as the system of record. Add a layer at Level 3 that runs the floor and talks to the ERP, so a job released in the ERP shows up on the dispatch list and a job finished on the floor shows up in the ERP.

This is not a new idea, and it is not mine. IMEC's own guidance on reporting is that a visibility tool "is not intended to replace ERP, accounting, CRM, or other core tools," those "can remain the systems of record," and the new layer "extends their value by helping users see related information together."^[4] In an earlier article the same center said it more bluntly: "You probably don't have to replace what you have. Just get them talking to each other."^[17]

What does the execution layer do? In plain terms:

• Dispatch. Each work center sees its queue in sequence, with the hot job on top because someone decided so.
• Status. Every job shows where it is, how long it has been there, and whether it is on hold.
• Notification. When a job stalls, the people who need to know are told. Nobody has to go find out.
• A shared picture. Scheduler, supervisor, GM, and sales all look at the same numbers at the same time.

That last point is the whole thing. The Lean Enterprise Institute defines visual management as putting "all tools, parts, production activities, and indicators of production system performance" in plain view "so the status of the system can be understood at a glance by everyone involved."^[18] A whiteboard does that for one cell. An execution layer does it for a building too big to see from one spot. It does not replace walking the floor. LEI defines a gemba walk as "grasping the current situation through direct observation and inquiry before taking action."^[19] The screen tells you where to walk. The walk tells you why.

I will say this once because it matters. One result I come back to came with no new system at all. A steel products maker I worked with had over $700,000 tied up in late orders. Six months later it was under $50,000. What changed was that everyone looked at the same numbers every day. The software's job is to make that habit easy to keep when the whiteboard stops scaling.

This is the system HixonOps builds first: an MES tied to the ERP and CRM you already have, deployed in your own cloud, with a flat license never priced per user. For the longer version of what an MES is and when a small shop is ready for one, read What is an MES, and does a small shop need one?

When you really do need a new ERP

Sometimes the ERP is the problem. NIST MEP success stories give three honest signals.

1. The vendor is sunsetting your product. Rathburn Tool, an Indiana CNC shop, faced the sunset of the ERP it had installed in 2012 and found it "very difficult to adequately research all of the ERP options available."^[20]
2. The ERP was built for lighter manufacturing than you do now. GAM Enterprises, a 32-person precision machining company in Illinois, recognized that its existing ERP was "one primarily designed for a lighter level of manufacturing" and went looking for one that would support material planning, production planning, and shop floor controls.^[21]
3. Your business runs on three systems that do not talk. MECCO, a Pennsylvania marking equipment maker, had "one for our CRM, one for our manufacturing, and one for our finance; all software is not really talking to one another."^[22]

If that is you, do what those shops did and go through your local MEP center before you talk to vendors. Purdue MEP helped Rathburn "map out their current processes and develop a vision for a future state," write an RFQ scope document, narrow the list, and sit through demos with an unbiased adviser in the room.^[20] IMEC started with GAM by "identifying all spreadsheets and tools currently in use, documenting all inputs and outputs for every step of order processing, production, and delivery."^[21] That documentation step is what tells you which layer is broken.

The short version

Your ERP is a Level 4 system. It keeps records, and it should keep doing that. The pain in your mornings is a Level 3 problem: execution, status, and who knows what. The standards have separated those layers since the mid-1990s.^[1] The audits and the research show that ERP replacement is slow and fails for management reasons a new vendor cannot fix.^[11][14] The MEP centers say to document what you have, start from the business question, and connect systems before you replace them.^[4][17]

So before you buy a new ERP, see your floor. If you find a records problem, replace the ERP with your MEP center's help. If you find an execution problem, which is what I found nearly everywhere, put an execution layer on the ERP you already own and get everyone looking at the same numbers.