Last winter I visited a metal fastener plant in Changhua. The owner had three vendors' decks spread across the conference table. Each one was a landscape feature comparison chart: inventory and order management, production work orders, cost analysis, mobile approvals, reporting center, ticked box by box. He pressed his finger on the middle one and said: this one has the most ticks, so we go with this one, right?
I didn't answer directly. First I walked to the back of the office and looked at the sales support staff's screen. Three windows were open: an Excel file for scheduling, a group chat in a messaging app, and a PDF order a customer had sent. She was on the phone while pushing a delivery date back two days in Excel, then turned to the group chat to let the team lead know. The systems actually running this plant were that Excel file and that group chat.
I'm Johnny Yang, Brightstream Technology, Taiwan. I've led implementation teams for seven years, 252+ projects across 32+ industries. So I'll put the conclusion up front: the most useless way for a traditional manufacturer to choose an ERP is to compare ticks on a feature chart. Every vendor's feature list looks about the same. That chart only tells you "what this system claims it can do." It tells you nothing about "where implementation will get stuck at your company." What actually decides success is three things: can the system tolerate the processes that are nonstandard, not pretty, but have made you money for twenty years; have the vendor's people actually stood on a floor in your industry, so they know why your material has losses and why your orders get add-ons; and five years from now, who is maintaining this thing, how many layers of people does it take to change one field, and is the consultant who understood you still around. If those three can't be answered, ten more ticks on the feature chart mean nothing.
Why feature-chart comparison is the most dangerous way to choose
To be fair, wanting to compare feature charts isn't strange. You have to make a decision, and you need something you can lay on the table, quantify, and justify to others. A feature chart looks exactly like that: neat, comparable, apparently objective. The instinct isn't wrong. The false sense of security is.
The problem is in three places. The feature chart is written by the vendor, and they own the definition of "has this feature." "Work order management" can mean full dispatching, job reporting, routings, and machine utilization. It can also mean a form where you type in a production order number. On paper it's the same phrase. On the actual screen, the work involved can differ tenfold.
Second, a feature chart compares "has it or not." But traditional-manufacturing implementations never blow up because "this feature doesn't exist." They blow up because "it exists, but it doesn't look the way you need it to." It has inventory management, but it assumes one part number has one unit. You make boards: you buy by the cai (a local board-volume unit), ship by the sheet, quote customers by the ping (about 3.3 m²), and the conversion changes with thickness. The feature chart still gets its tick.
The worst is the third point: a feature chart puts every vendor on the same flat plane, which means giving up the variable that actually separates them. Two vendors with the same number of ticks; one vendor's consultant has done five projects in your industry, the other has done none. You can't see that anywhere in the chart. And in the implementations I've seen, that variable affects the outcome far more than any feature difference. I covered the full sequence for judging whether a vendor fits you in how to choose an ERP vendor. Here I'll only cover the three criteria traditional manufacturers trip on.
Criterion 1: can the system tolerate your nonstandard processes
This is the most counterintuitive one, and the one you're most easily talked out of.
During the demo, the consultant will tell you: the way you do this isn't standard, and implementation will standardize it. That's part of the value of implementation. Half of that is true. You do have a pile of processes left over from history, and changing them would be better. The other half is dangerous: traditional manufacturers have a class of processes that look nonstandard but are actually the core competitive advantage of the business.
Two types I've seen on the floor. One is a machine shop making custom-built parts. When the customer orders, the drawing isn't final yet, so sales opens a "provisional order" so the floor can start preparing material. Once the drawing is final, they go back and change specs, quantities, and delivery dates. Standard ERP logic says production can only start once the order is confirmed, and once the work order is open, the BOM is locked. Follow the standard flow, and sales has to wait for the final drawing before ordering. Material prep slips a week, and the customer walks.
The other is "rush orders scheduled by someone's head." The plant manager glances at the machines, glances at which technicians showed up today, glances at which customer chases hardest, and has the sequence done in five minutes. That judgment contains a lot that can't be written as rules. Bring in a scheduling module, and the sequence it produces is optimal on paper. On the floor, the plant manager will simply ignore it.
So the right way to ask isn't "can you customize," because every vendor will say yes. Ask this: does this way of working go through a standard path in your system, or does it need code changes? If it needs code changes, will that part get overwritten in future version updates?
Ask that, and you'll hear the difference immediately. Some will configure it in the system right in front of you. Some will say they need to evaluate. Some will say "actually, your process could be adjusted a bit." That last one isn't a bad answer, but follow up: after the adjustment, who's responsible for the material prep on that provisional order? If they can't answer, they haven't really understood your floor.
As for what to keep and what to change while you're at it, the test is one sentence: if you remove this nonstandard practice, will you make money more slowly? If yes, it's your core, and the system has to work around it. If no, it's just habit. Change it once, during implementation.
Criterion 2: does the vendor understand the floor in your industry
This is the easiest one to verify, and the one most people skip.
Most owners ask "have you done projects in our industry." The question is right, but it's too weak. They'll always say yes and bring out a client list. A client list only proves they signed contracts. It doesn't prove the person sitting across from you understands your floor. About 80% of implementation quality depends on the consultant who actually comes to your company.
Change the question from "have you done this" to "explain it to me." Pick the most troublesome detail in your industry and have them describe it. If you do plastic injection molding, ask: how is scrap during mold trials usually accounted for? If you do metal machining, ask: while material is out for heat treatment at a subcontractor, whose books is it on? People who know will go straight into details, and even ask you back, "do you send every batch out, or only the large parts?" People who don't will give you a polished, empty speech about systems, management, process optimization, and never land on the point you asked about.
One more practical thing: you need to meet the person who will actually work on your project, not the salesperson. A sharp team before signing, and a new hire sent as the consultant after signing: I've seen that more than once. Before you sign, ask directly: who is the consultant on this project, how many projects in our industry have they done, can I talk to them for thirty minutes first? Vendors willing to arrange that usually aren't bad. If they push it off with one excuse after another, take note.
Another method few people use, but which works: walk them through the plant. To the receiving area, to the corner where work-in-progress sits, to the sales support staff's desk. People who really understand will stop along the way to ask questions, and crouch down to look at the handwritten labels taped to the shelves. People who don't will nod the whole way and compliment how well-managed you are.
Criterion 3: who maintains it five years from now
This is the one most easily overlooked, and the one that does the most damage. Traditional-manufacturing owners judge a system by "is it easy to use once it's in," but implementation is only a small part of an ERP's life. The rest is maintenance.
Picture this: three years from now you need to add a shipping label format a new customer requires. Five years from now you take on a new product line that needs one more process step. Or your accountant retires, and the replacement needs to change how a set of reports is calculated. When these things happen, who touches the system?
How many layers of people does one change go through, and roughly how long does it take. This isn't a question about service attitude. It's about structure. Some vendors' consultants can adjust things themselves in system settings. Others need a requirements ticket, a development slot, and a version release for every small change. You can't feel the difference during implementation. By the second year after go-live, it's glaring. When changing one field takes six weeks, your people stop trying to change the system and go back to Excel.
Is the person who understood you still there, and what happens when they're replaced. Consultant turnover is normal. You can't expect the same person to stay with you for ten years. What you need to ask is how they hand over: the custom logic, why it was designed that way, whether there's documentation. Some companies keep full project records, so whoever takes over can understand them. Others keep it all in one person's head. When that person leaves, you're an orphan.
How will your customizations be handled in future version updates. You've changed a dozen-plus things. Five years from now the system needs an upgrade. Do those carry over automatically or need to be rebuilt? Who pays for the rebuild? Ask the other question while you're at it: if you switch systems one day, can your part numbers, customers, and five years of transaction records be exported in a standard format? Nobody likes discussing divorce before the wedding, but if the answer is vague or they look uncomfortable, be especially careful.
The core of this set of questions: you're not buying software. You're entering a long-term relationship. The succession timeline in traditional manufacturing makes it more complicated. Many plants implementing systems now are really doing it so the next generation can take over. If that's your situation, look at one more thing: can the knowledge in this system actually be handed over. I covered that more fully in succession and digitization.
How to ask these three questions, and when you shouldn't buy
I've put it in a table. Bring it to your next vendor meeting and ask directly. What matters isn't the questions themselves. It's how they answer.
| Criterion | What to ask | A good answer looks like | Answers that should worry you |
|---|---|---|---|
| Tolerates nonstandard processes | Does this way of working go through a standard path in your system, or does it need code changes? Will it get overwritten in version updates? | Opens the system and demonstrates on the spot, or says clearly "this part needs custom work, and we'll preserve it during updates" | "We can do anything" or "that's easy," but can't explain how |
| Understands the floor | The most troublesome part of our industry is X. Tell me how you usually handle it | Starts into details and asks how your company does it | Talks about systems and management benefits, never touches the point you asked about |
| Who maintains it in five years | After go-live, how many layers of people does changing one field go through? How do you hand over when the consultant changes? | Can describe a concrete process and timeline, willing to discuss handover documentation | "Just come to us with any problem" — a promise with no structure |
If all three answers are concrete, the vendor makes the final round. If they clearly dodge one, that doesn't rule them out, but write that risk into the contract or acceptance criteria. If they sidestep all three, don't buy, no matter how full their feature chart is.
When you shouldn't buy at all right now
This part works against anyone trying to sell a system, but I'll say it anyway. If this is you right now: every salesperson handles orders differently, part numbers have no consistent coding, inventory counts have never been accurate, and nobody has time to own this project — then implementing an ERP now will just move the mess, untouched, into a place that's harder to change. Six months later everyone is back on Excel.
What you should do is clean up the most painful process first, usually either inventory or orders, get it running smoothly with the tools you already have, and assign one person to truly own it. That person doesn't need an IT background, but they must carry weight inside the plant, and the owner must be willing to take part of their regular work off their plate. The other situation where you shouldn't rush: the owner hasn't figured out what problem to solve, and it's only because a competitor did it, or the second generation came back and felt it should be done. Buy a system then, and requirements will keep growing during the project. Timeline and scope will both get out of control.
So what you should do now
Before you book the next vendor, spend two hours on three things.
First, write down the three "most nonstandard" processes in your company. They don't have to be well-written. Write the ones you feel a bit sheepish saying out loud — provisional orders, delivery dates changed by word of mouth, parameters the technicians adjust by experience. Then mark each one with a single word: keep, or change. The test is the sentence from earlier: would removing it make you money more slowly? That sheet of paper is the backbone of every meeting from here on.
Second, prepare an "insider question" for your industry, the kind of detail an outsider definitely can't answer. Ask every vendor the same question. You'll feel the difference in answer quality very clearly.
Third, take them on a walk through the floor. Don't let the evaluation happen only in the conference room. Walk them to that shelf covered in handwritten labels and see whether they stop.
Once you've done these three, you're no longer holding just a feature comparison chart. You have three criteria that actually separate vendors. If you want someone to work through the first one with you, Brightstream offers a free 30-minute diagnostic. We can start by talking through whether those three nonstandard processes should be kept or changed.




