Every ERP shortlist looks reassuring on paper. The platforms demo well, the reference calls check out, and the feature grids start to converge. Then the system meets a live plant, and the gaps that no demo revealed begin to cost money.

That is the pattern behind most disappointing ERP outcomes. The platform did not fail because the software could not work. It failed because it was never tested against the conditions a plant actually imposes, and because the questions that predict that behavior are rarely the ones on a standard evaluation grid.

The six questions below are worth answering before choosing an ERP platform, ideally before a shortlist even forms. They are not about which vendor looks strongest in a demo. They are about how a platform behaves when transaction volume spikes, when a divestiture sets the clock, and when AI needs data it can actually use.

Quick answer: Before choosing an ERP platform, manufacturers should answer six questions. Will the platform fit the plant, or will the plant fit the platform? How much of your operation still runs outside the ERP today? How fast can the first plant actually go live? What happens the day transaction volume spikes? Can the system act, or only report? And has your partner done this in a live plant before?

Key Takeaways

• The questions that predict an ERP outcome are about production behavior, not demo performance.
• A horizontal platform reaches manufacturing fit through customization; a manufacturing native platform inherits it, which changes cost, speed, and upgrade risk.
• The spreadsheets and emailed files running real decisions today are the true starting point of any ERP project.
• Time to value should be measured by when the first plant goes live, not when the full program finishes.
• When Tenneco’s braking division answered questions like these, a standard S/4HANA migration was ruled out and QAD Adaptive deployed across 8 plants and 3 continents in 18 months.

1. Will the Platform Fit the Plant, or Will the Plant Have to Fit the Platform?

Horizontal platforms reach manufacturing fit through configuration and custom code, so the plant bends to the software. A manufacturing native platform such as QAD Adaptive is built around production, so fit is inherited rather than built. The direction that fit runs decides how much you customize now and how much you defend at every upgrade.

General purpose platforms like SAP and Oracle serve many industries, and manufacturing is one of them. To reflect how a specific plant sequences work, plans materials, and manages quality, they usually have to be configured heavily and extended with custom code. That is not a defect. It is what a general system does to fit a specific plant.

A manufacturing native platform starts from the plant. QAD Adaptive supports discrete, repetitive, mixed mode, batch, and lean manufacturing, with MRP, MPS, shop floor control, and quality management as standard. The practical test: are you selecting standard functionality, or are you funding a rebuild of what a purpose built system already includes?

2. How Much of Your Operation Still Runs Outside the ERP Today?

Count the spreadsheets, emailed files, and manual approvals that carry real operational decisions. That shadow system is the honest starting point of any ERP project. A platform that cannot absorb it will simply push the workarounds into a newer wrapper.

When planning logic, approvals, and inventory decisions live outside the ERP, it is usually the clearest sign the current architecture no longer supports the business. Bringing those workflows in, often called application rationalization, is a prerequisite for both clean operations and any future AI, not an afterthought.

Tenneco’s braking division is a vivid example. Despite naming SAP as its ERP, it ran 18 separate “solutions” in Excel, with MRP on manually updated sheets. If a platform selection ignores that shadow system, the project pays a premium to recreate old constraints inside new software.

3. How Fast Can the First Plant Actually Go Live?

Time to value is not the length of the whole program. It is how soon a single plant is running and capturing improvement. A three month divestiture deadline and a three year roadmap call for completely different strategies, and long migrations expose you to market shifts before the first plant ever goes live.

The default big bang model, one platform, all sites, full alignment before go live, treats every situation the same way. But three to five years into a large migration, markets shift and the strategy you designed the system around may no longer apply.

“Nobody has 10 years,” Sam Gupta, CEO of Elevate IQ, said during a recent QAD webinar. “Nobody knows what is going to happen in the next 10 years.” When speed matters more than consolidation, a purpose built platform or a two tier plant deployment delivers value long before a full migration would.

4. What Happens the Day Transaction Volume Spikes?

Demos run on clean, low volume data. Production does not. The real question is whether the platform holds when sequencing tightens, volumes surge, and decisions have to be made in minutes rather than hours. Architecture becomes visible under load, not in a sandbox.

Manufacturing imposes constraints that a demo never shows: tight sequencing and dependency between processes, high transaction volume with no tolerance for latency, and little patience for a fix it later approach. A platform that looks capable in a controlled environment can still stall when real automation feeds real time data.

Cloud native architecture helps here, with continuous updates and no dependence on nightly batch processing to keep decisions current. Ask a prospective platform to show you production, not a demo. The two are not the same conversation.

5. Can the System Act, or Only Report?

A system of record tells you what happened after month end, when the window to intervene has closed. A system of action tells you what to do now, before the shortage, before the scrap accumulates, before the shipment misses. Agentic AI enables that shift, but only on trusted, connected data.

This is why the data question and the AI question are the same question. Embedded AI can surface supply chain risks before they become shortages, flag MRP exceptions before they become delays, and recommend the next best action without waiting for someone to run a report. None of that works if the underlying data is fragmented across spreadsheets and disconnected systems.

So the honest version is not “does the platform have AI?” It is “is our data in a state where AI could do anything useful with it, and does this platform keep it that way?”

6. Has Your Partner Done This in a Live Plant Before?

Platforms do not earn production trust. Execution does. The right partner has managed data migration, cutover, and operational readiness in a live plant, not just presented in a boardroom, and treats each of those as a risk surface to manage rather than a checklist item to clear.

A platform that fits perfectly on paper can still fail if the team implementing it understands software but not manufacturing reality. ERP success depends as much on the execution partner as on the technology.

The real ERP decision is not only about business process requirements. It is about master data, integration architecture, data flow, talent bandwidth, and operational continuity during the transition. Evaluate that partner experience with the same rigor you apply to the platform itself.

What Happens When a Manufacturer Answers These Honestly

When Tenneco’s braking division put its ERP decision through questions like these, the answers ruled out a standard S/4HANA migration. It would have cost roughly double and could not meet the 16 to 18 month divestiture timeline. QAD Adaptive was chosen instead for cost, timeline fit, and purpose built automotive functionality that did not have to be built from scratch.

Across 8 plants and 3 continents in 18 months, MRP moved fully into the ERP with no spreadsheet uploads or downloads, and shipping went from six people working multiple approvals to one person following a two step process. Eight plants. Three continents. Eighteen months.

The framework did not point to a brand. It pointed to a category and a plan that could survive production.

Start With the Questions, Not the Shortlist

Feature comparisons feel like the substance of an ERP decision. More often they are the surface. The answers that actually predict how a platform performs come from these six questions about fit, workflow debt, speed, load, action, and execution.

Answer them candidly and the direction usually becomes clearer than any vendor demo could make it.

For readers ready to go deeper, Arista’s guide, “Big ERP Isn’t the Only Option” walks through the alternatives to a horizontal migration and pairs with a decision checklist you can bring to your own team.

FAQ

What questions should manufacturers ask before choosing an ERP platform?

Six questions matter most: will the platform fit the plant or will the plant fit the platform, how much of the operation still runs outside the ERP today, how fast can the first plant go live, what happens when transaction volume spikes, can the system act or only report, and has the implementation partner done this in a live plant before.

What is the most overlooked question when selecting a manufacturing ERP?

How much of the operation still runs outside the ERP. The spreadsheets, emailed files, and manual approvals carrying real decisions are the true starting point. A platform that cannot absorb that shadow system just moves the workarounds into newer software, so rationalizing those workflows is a prerequisite, not an afterthought.

Why do ERP platforms fail after go live?

Most do not fail because the software cannot work. They fail because the platform was never tested against production constraints such as tight sequencing, high transaction volume, and decisions that must be made in minutes. Execution discipline around data migration, cutover, and readiness is what earns production trust.

How is choosing a manufacturing ERP different from general ERP selection?

Manufacturing adds constraints that generic evaluation grids miss: production sequencing, shop floor and MRP fit, tolerance for transaction load, and clean data for AI. A manufacturing native platform such as QAD Adaptive is built around these, reducing customization, cost, and upgrade risk compared with a horizontal platform.

Can cloud ERP be deployed without disrupting live production?

Yes. With disciplined architecture, thorough cutover and data migration planning, and a partner experienced in manufacturing, cloud ERP can go live at an active plant without interrupting operations, as shown when QAD Adaptive deployed across 8 plants and 3 continents in 18 months.