Should your ERP be custom-built? The SME sweet spot and the most expensive mistake

The most expensive mistake with custom-built work isn't the money. It's moving your old bad process into a new system unchanged: not a minute of work time saved, and one more system to maintain. This post covers how to find the sweet spot, and when you don't need custom-built work at all.

Johnny Yang16 min read
Should your ERP be custom-built? The SME sweet spot and the most expensive mistake

"Johnny, our company's processes are really unique. Off-the-shelf software won't fit. It has to be custom-built, right?"

I hear this about thirty times a year. Every time, I ask one question back first: "When you say 'unique', is it actually unique, or is it just 'the way we've always done it'?"

They usually freeze for three seconds.

In this post I want to talk honestly about custom-built software. I don't sell courses. I won't tell you "only fully custom-built counts as professional," and I won't tell you "packaged software is cheapest, just make do." I've led my team through ERP implementations for seven years: 252+ projects across 32+ industries, from traditional factories and trading companies to e-commerce and service businesses. After seeing that many, here's what I found: The biggest mistake SMEs make with custom-built work is not spending too much on it. It's paying to move a process that was already bad into a new system, unchanged.

Conclusion first. If you're short on time, this section is enough:

The point of custom-built work is not to "move" your current process over. It's to use the rare chance an implementation gives you to "fix" the process. You spend serious money on a new system, and the whole company grinds through three to six months with you. You get only a handful of chances like that in a lifetime. If you just copy that winding Excel process one-to-one into the new system, you've paid for custom-built work, you're maintaining a new system, you pay license fees every year, and your staff hasn't saved a single minute. They've just gone from "filling in Excel" to "filling in the new system."

This is called "digitizing a bad process." It's the most expensive, most invisible, and most common waste in the entire implementation.

1. The most expensive mistake: moving as-is turns "filling in Excel" into "filling in the new system"

Let me describe a very typical scene. Check whether your company looks like this.

Before one trading company came to me, they went to a software vendor with a very attractive quotation. The vendor's salesperson was smooth: "No problem. However you run things now, we'll custom-build the system to match exactly. Zero learning curve for your staff, painless go-live!" The owner felt completely reassured. This vendor gets me, he thought.

The result? At that company, a single purchase order was entered once by purchasing, copied by the sales support staff into another sheet to be checked, then keyed in again by accounting into a third file for reconciliation. The same numbers, entered by three people, once each. The vendor "obediently" turned those three steps into three screens in the system, unchanged.

After go-live, staff went from "entering it three times in Excel" to "entering it three times in the new system." The same number still got typed by hand three times. The interface just looked nicer. The owner asked: "Why isn't it any faster?" The vendor said: "Because that's how your process already works. We built what you asked for."

That answer isn't wrong. But it's exactly the problem.

The vendor moved your overcomplicated bad process into the new system unchanged. You're just suffering in a different place. Not a minute of work time saved, plus one more system to maintain and one more license fee to pay. You thought you bought "efficiency." What you actually bought was "a premium replica of the same inefficiency."

What should have happened? Before the purchase order was built into the system, someone should have stood up and asked: "Why are three people each entering this number? Can't the system automatically carry the purchase order data into the check sheet and the reconciliation statement? Enter it once, everything downstream fills in automatically. Done."

Ask that once, and three steps become one. The same document, which used to cost three people five minutes each, becomes one person entering it and everyone else clicking "confirm" in the system. That is implementation. That is the value custom-built work is supposed to create.

Where's the difference? The difference is whether anyone is willing to "cut the process" for you before building anything.

Bad vendors take every request as given. Whatever you ask for, you get, because taking requirements as-is saves them effort, you pay, they collect, and everyone's happy (on paper). Good vendors frown first and ask you a pile of uncomfortable questions: Why is this entered twice? Is this sign-off pure ritual, where the person signing doesn't even read it? Can this data be carried over automatically from the previous step?

When the cutting is done, only the "truly necessary" process is left to digitize. Only then is every dollar you spend on custom-built work spent where it counts.

On why implementations so often fail, with the money spent and nothing to show for it, I go into more detail in Why ERP implementations fail. Many failures trace back to this root: the system isn't bad; the bad habits got moved in along with it.

2. The price of more custom-built work: more expensive, harder to upgrade, more locked in

OK. Say you really have cut your processes, and you do have non-standard needs that must be custom-built. I still want you to see the "three costs" of custom-built work clearly first. Custom-built work feels great now and hurts later.

Cost one: the more custom-built, the more expensive — and it's an "ongoing" expense.

Custom-built work is not a one-time fee. You think once the custom development line on the quotation is paid, you're done? No. Every piece of custom code needs someone to maintain it from then on. The system needs adjusting, connecting to new things, bug fixes. The vendor maintains standard features for hundreds of customers at once, so your share is cheap. Your custom-built code is "one of a kind," so you carry the whole maintenance cost yourself. The more custom-built work you have, the thicker your annual maintenance contract gets. I've laid out these long-term costs hidden behind the quotation in How to calculate ERP implementation cost. Read it before you sign anything.

Cost two: the more custom-built, the harder to upgrade.

This is the one most people ignore, and the most damaging. ERP vendors release new versions every so often — patching security holes, adding features, keeping up with new government regulations (like changes to Taiwan's e-invoice format). Customers on standard features upgrade with one click. But if you've custom-built a pile of things, every time the vendor releases a new version, your custom code may "collide" with it and conflict. At best you pay the vendor to readjust it. At worst you simply can't upgrade, stuck on the old version, watching new features that have nothing to do with you.

I've seen companies custom-build so much that they didn't dare touch the system for five years, because touching it would bring the whole thing down. That isn't an asset. It's a burden.

Cost three: the more custom-built, the more locked in to the vendor.

The deeper and more unusual your custom-built system gets, the more it becomes something only the original vendor in the whole world understands. The day you're unhappy with that vendor and want to switch or negotiate, you'll find you can't leave — no one else can take on that pile of custom code, and a new team would need three months just to understand it. At the negotiating table, you're the side with no leverage. The vendor raises maintenance fees, service gets worse, and you just put up with it.

I'm not saying custom-built work is always bad. I'm saying be clear about this: every extra piece of custom-built work is one more rope tied around yourself. Tie it when it should be tied. When it shouldn't, don't tie it on casually just because "custom-built fits better anyway."

3. How to find the sweet spot: cut the process first, use standard where you can, custom-build only your core differentiator

So how do you actually find the "sweet spot"? Here's a very practical three-step method. The order cannot be reversed.

Step one: cut the process first, then talk about digitizing.

This is the core of the whole post, so I'll say it again: Real custom-built work means using the rare chance an implementation gives you to "fix" the process, not "move" it over. Before implementation, lay your current process out on the table and ask three questions at every step:

  • Why is this step entering the data a second time? (Duplicate entry → cut it, carry it over automatically)
  • Is this checkpoint actually controlling anything, or is it a ritual that exists because "it's always been there"? (Pure ritual → cut it)
  • Can this data be carried over automatically from the previous step? (If it can be automatic → don't make a person do it by hand)

Once you've cut, you'll find that half of the special needs you thought "had to be custom-built" were just bad habits. Cut them and they're gone. The remaining half are the real needs.

Step two: if a standard feature works, use the standard feature.

With the real needs left after cutting, don't rush to custom-build. First check whether the ERP's "standard features" can cover them. Today's mid-to-large ERPs cover over 80% of common scenarios with standard features — inventory, purchasing-sales-inventory, approvals, reports, mostly out of the box. The benefits of standard features were covered above: cheap, easy to upgrade, no lock-in. Even if a standard feature forces you to "adjust how you work a little," as long as the adjustment is reasonable, you should change the people to fit the system, not change the system to fit old habits.

Remember one line: If a standard feature can solve it, don't custom-build it. Save custom-built work for the places where "no standard feature can really solve it."

Step three: custom-build only your "core differentiator."

So what's worth custom-building? The core your company makes its living on, the part that's different from everyone else and actually creates your competitive edge.

Example: if you're a factory making special-order products, and your "quotation logic" and "BOM explosion method" are know-how nobody can copy, then custom-building that part is worth it — it's how you make money. But your "leave request process" or "general office supply purchasing," things that are about the same at every company, just use the standard features. Don't custom-build them.

The sweet spot formula is simple: Fix the process first → use standard wherever standard works → spend money only on custom-building your real core differentiator. Spend willingly where you're different, don't waste money where you're the same as everyone else. That's the sweet spot.

If you can't yet tell which parts of your company are the "core differentiator" and which are "generic processes," read The SME ERP buying guide first. Get a clear inventory of your needs, then decide whether to custom-build.

4. Honestly, you may not need custom-built work at all

After the first three sections, here's a truth many vendors won't tell you — because telling you means they earn less:

A lot of the time, you don't need to overhaul the system at all. Connect your existing systems through APIs, add automated workflows, and the problem is solved.

This is where Brightstream differs most from a lot of our competitors. When others hear "special requirement," their instinct is to quote a custom development fee. Our instinct is to ask first: "Does this really require touching the ERP itself? Or can we just 'connect' a few systems?"

Here are a few real scenarios:

Scenario one: your e-commerce orders get copied into the ERP by hand. Many owners want to custom-build the ERP with "built-in e-commerce features," and the quotation is terrifying. You don't need it — connect the e-commerce platform to the ERP through an API, and the moment an order is placed it flows into the ERP automatically and inventory is deducted in real time, with no one touching it. I covered the full approach to this kind of cross-channel inventory integration in How to do omnichannel inventory integration. It costs far less than custom-building the ERP.

Scenario two: someone compiles a report for the owner by hand every day. No need to custom-build a reporting module. Set up an automated workflow that pulls the data every morning, does the calculations, and sends it to the owner's inbox. Not one line of custom code.

Scenario three: after a customer orders, you need to notify sales, open a work order, and schedule production automatically. Chain these actions with an automated workflow. It's cheaper and more flexible than "custom-building one big module." When the process changes later, you adjust the workflow, without going back into the ERP's core.

Since AI arrived in the last few years, the range of what can be automated has grown again. I've listed what you can now hand to the system to run on its own in What an AI-powered ERP can do. You'll find that many tasks you thought "needed a person" can now be automated.

Don't custom-build for the sake of custom-building. Custom-built work is the last resort, not the first option. Before touching the ERP core, ask one question: can integration and automation solve this? If yes, it's cheaper, there's no lock-in, it's easier to maintain. Why wouldn't you?

Push that question one step further and you get another one owners ask a lot: if we're changing this much anyway, why not just build our own system? When that idea holds up and when it turns into a bottomless pit, I've written up as a decision framework in Build or buy? A decision framework for SMEs.

As for the e-commerce integration in scenario one, in practice there's a pile of details on order formats, returns, free gifts, and multiple warehouses to sort out first. I wrote that up separately in How do multi-platform e-commerce orders get into your ERP? What to think through before integrating.

Quick comparison: bad custom-built work vs good custom-built work

Can't tell which kind your vendor is doing for you? Check this table:

AspectBad custom-built work (digitizing a bad process)Good custom-built work (fixing the process along the way)
Before buildingBuilds your current process exactly as-isCuts the process first, questions every step
Duplicate entryKept; still entered three times in the systemCut; entered once, carried over automatically
Ritual sign-offsKept; you asked for it, so it's builtPulled out and questioned: "does this checkpoint mean anything?"
Work timeBarely saved; just a new interfaceActually saved; people moved to more valuable work
Standard featuresAvoided where possible; loves custom-builtUsed wherever possible; custom-built saved for differentiators
Fee logicThe more you want, the more I earnSaves you unnecessary custom-built work first
Upgrades/futureHarder to upgrade, more locked in over timeCustom-built work minimized, upgrade flexibility kept
Relationship with youTighter and tighter; you can't leaveYou can always switch or negotiate

Hold the vendor you're talking to up against this table and it's obvious. If you still can't tell which kind they are, use it together with the checklist in How to choose an ERP vendor.

Before you sign, ask this one question

If you remember only one thing from this post, remember this. Before signing, in front of the salesperson, ask directly:

"Will you 'cut my process' first, or build my current process 'exactly as-is'?"

Then watch how they react.

If they say: "Of course we build what you need. You know your company best, we'll follow your lead." — it sounds considerate, but it's a warning sign. It means they're ready to digitize your bad process. Whatever you ask for, you get. No questioning, no gatekeeping, because that's easier for them.

If instead they start asking you questions — "Why does this step need two signatures?" "Can this data carry over automatically? Why copy it by hand?" "What's the actual purpose of this step?" — congratulations, you've found a vendor that will cut your process. If their questions annoy you a bit, even offend you a bit, good. That means they're actually thinking for you.

A good vendor will always make you uncomfortably examine your own process before building anything. That discomfort is where the next several years of waste get saved.

FAQ

Q: Our processes really are unique and have to be custom-built. Can't that be left alone? "Unique" and "complicated" are two different things. Your real differentiator (say, proprietary quotation logic or manufacturing process) is unique. It should be custom-built and kept. But a lot of what owners call "unique," once you take it apart, is an accumulation of "duplicate entry," "ritual sign-offs," and "because a certain senior employee likes doing it this way." That isn't unique. It's baggage. Have the vendor cut first. What's left after cutting is what really has to be custom-built. Usually, after cutting, it's much less than you'd expect.

Q: Fully packaged, zero custom-built — isn't that the cheapest? Not necessarily. Packaged software is the cheapest and easiest to upgrade, that's true. But if it forces you to make do and tie your own hands on your "real core differentiator," the custom-built fees you saved get paid back double in lost competitiveness. The sweet spot isn't "fully packaged" or "fully custom-built." It's standard where things are generic, custom-built only where your core is. Save where you should save, spend where you should spend.

Q: What share of the implementation budget is reasonable for custom-built work? This varies a lot by industry and by need. I won't give you a fake precise number to make you feel better. But here's a rule of thumb: if custom-built fees "far exceed" the software license fee itself, be very alert — it usually means either you picked the wrong system (its standard features are too far from what you need), or the vendor is moving your bad process over as-is. That's when you stop and take inventory again, not grit your teeth and pay. For the detailed cost structure, see How to calculate ERP implementation cost.

Q: We've already gone live and only now realized it's "a digitized bad process." Can it be saved? It can, but it takes more work than doing it right from the start. The approach: go back and inventory which steps are duplicate entry or pure ritual, then see whether API integration and automated workflows can "patch" them — often you don't need to tear it down and start over. Automate a few bottlenecks and you can win back most of the lost work time. For core parts that truly can't be touched, then evaluate partial rebuilds. The key is to stop making do. The waste piles up every day.


If you've read this far and you're unsure whether to go custom-built, or you suspect the quotation in your hands is "moving your bad process over," come talk to me. Brightstream offers a free 30-minute diagnostic — just pick a time slot online to book. I'll look at it directly with you: which processes should be cut, where standard features are enough, and where API integration and automation can solve it without spending big on custom-built work. No sales pitch, no courses. Just making sure your money goes where it counts — because fixing the process is always far cheaper than moving it over.

Found this useful? Pass it on
Prices published · free range estimate

So what does a build to your process actually cost?

From NT$300,000, priced against your spec, with same-scope changes free during the maintenance period. The prices and the estimator are in the first screen — get a range first, you do not have to decide anything yet.

Written scope, then a fixed price · staged acceptance · the code and the data are yours · or just book a free 30-minute diagnostic

Johnny Yang (熊董)
Johnny Yang (熊董)
Founder, Brightstream Technology

Serial founder leading a small team that builds systems into an owner's everyday work.

As a founder
7 yrs
Projects delivered
252+
Industries served
32+
About Johnny