Criteria for choosing an ERP: the questions to ask during a demo

ERP selection usually starts with the wrong question: “which ERP is best?” That question has no answer, because the criteria change with the type of production. The right question is: in our way of manufacturing, which decisions will come out of the system, and can this system hold the data those decisions need?

The wrong question, the right question

Comparing feature lists is the most time-consuming and least discriminating part of a selection process. Above a certain level of maturity, feature lists resemble each other; the difference is not in the list but in how the feature behaves in your way of manufacturing.

The discriminating sequence is: which decision do we expect from the system, which data does that decision rest on, and how and by whom is that data entered on the floor? With clear answers to those three, choosing the right system becomes easy. Without them, even the broadest system turns into a record store.

Selecting a system is not a software decision but a commitment about which decisions will rest on data.

Scope: how far does one system reach

The first criterion is how far the system reaches. Accounting and inventory exist in almost every system; the divergence begins inside production. Are shop-floor records, downtime, scrap, quality and actual cost held in the same data model, or attached to a separate product?

This is an architectural choice, not a quality indicator — there are cases where either approach is valid. But if separate products are involved, it must be clear from the outset who owns the mapping between the two systems. A notable share of the cost deviations we see on site comes from an incomplete work-order mapping between ERP and MES .

Adaptability and the open box

No manufacturing plant maps perfectly onto a standard installation. The second criterion is therefore not how far the system can be adapted but how adaptation is done. If it happens at screen and field level, independent of version upgrades, the system stays alive. If it requires touching core code, every upgrade becomes a new project.

This is why we care about the “open box” approach: the data model and business rules are visible, and adaptation happens through defined extension points. It does not mean unlimited customisation — on the contrary, it means knowing from the start how far adaptation can go. We set out the scope on the ERP solution page .

Integration: the system's boundaries

The third criterion is the system's relationship with the world outside it. In manufacturing that runs in three directions: field equipment below, existing applications alongside, statutory obligations above.

• The field side. Counters, PLCs and operator terminals produce different kinds of data. Deciding which data is read from which layer — otomasyon entegrasyonu — is the item that causes most project delay; always ask for a live example during the demo.
• Adjacent systems. How it maps to existing accounting, payroll or customer management applications: file transfer, service-based connection, or double entry?
• Obligations. E-document processes and, if your sector is regulated, audit trail requirements. Adding an audit trail afterwards is one of the most expensive retrofits there is.

Support and continuity model

The fourth criterion is usually the least asked and the longest lived. Implementation takes a few months; the support relationship lasts years. What to look at is not the response-time commitment — every contract states one — but whether the person solving the problem understands manufacturing processes.

A concrete measure: how many people on the support team have implemented in your type of production? At the next version upgrade, what happens to your adaptations and who carries them over? If that answer is vague, the upgrade never really happens and the system ages out within five years.

Ten questions to ask during a demo

A demo is a rehearsed scenario. The revealing information appears when you step outside it. The questions we have seen work best:

1. 1Can you enter one example from our own bill of materials live and calculate its cost?
2. 2On which screen do we see actual cost when a work order is thirty percent complete?
3. 3Are setup scrap and run scrap held separately, and how does small-batch cost come out?
4. 4Do we define the downtime reason tree ourselves, or does a fixed list come with the system?
5. 5Starting from a customer complaint, in how many steps do we reach which raw material was used in which batch?
6. 6If an alternative is defined for a critical material, how does MRP evaluate it?
7. 7What happens to an added field at a version upgrade — is it carried over or redone by hand?
8. 8Can we adapt this screen ourselves, or does every change require raising a request?
9. 9If data from a shop-floor terminal arrives late, what does the system do and how is double entry prevented?
10. 10Can we speak to a reference that implemented in our type of production?

The tenth question yields more than the other nine combined. And in that reference call there is one question to ask: if you went back, what would you do differently?

The invisible part of the cost

The quoted figure is one part of total cost. The items that determine total cost of ownership over the years usually do not appear as separate lines in the quote.

Over five years the last two items can exceed the first. That makes the method of adaptation a more decisive criterion at selection time than price. We describe our approach to scaling on the scalability page .

Frequently asked questions

An industry-specific system or a general ERP?

The distinction is in the data model, not the industry label. In packaging, if concepts such as square-metre-to-unit conversion, dies and setup scrap are defined in the data model, the system fits the industry; if the industry name appears only in the interface, it does not. Testing with your own bill of materials during the demo shows the difference within an hour.

Should ERP and MES come from the same system?

Not necessarily, but work order, batch and cost must mean the same thing in both systems. If separate products are chosen, ownership and maintenance of that mapping must be explicit in the contract; the most common deviations occur exactly at that boundary.

How long does implementation take?

What sets the timeline is not the software but the state of existing data and the speed of decision-making. In a plant with current bill-of-materials and inventory data, the core scope goes live within a few months; if data has to be compiled, it takes longer. An unrealistic timeline is met by cutting data cleanup — and the cost of that surfaces later.

At which stage should shop-floor recording start?

Immediately after inventory and the bill of materials work correctly. Without shop-floor records actual cost cannot be calculated; but if the bill of materials is wrong, shop-floor records only produce the wrong cost faster. That is why the order matters.