Brightstream Technology · custom ERP implementation
Commissioning an ERP —
what does it actually cost?
From NT$300,000, priced against your spec.Maintenance from NT$15,000 a month after launch; no headcount surcharge up to 200 people, above 200 quoted separately.
The price is on the first line because it should have been said first.
But if you have failed at this once, the question is not how much —it is what makes this time different.
Johnny Yang talks to you · no salespeople · we choose people, not industries
Get a range first
Two minutes to fill in, and we come back with the range a spec like yours usually lands in, and which parts would send the price up. Nothing to decide yet.
First, the thing you are actually afraid of
What you fear is not the price,
isfail a second time。
Owners who get as far as commissioning a custom build are usually not doing it for the first time. The money is survivable; a year and the team's confidence are not. You have probably hit at least one of the three below.
Extras start appearing halfway
Signed on lovely terms, then halfway through "that is out of scope". The budget goes from NT$500,000 to NT$900,000 and you are in no position to refuse, because stopping wastes everything already done.
Six months in and nothing to look at
You paid, you assigned people, and every progress check comes back "still in development". By the time you see a screen it is six months later and nothing like what you had in mind.
Abandoned after launch
The salesperson disappears after acceptance. Change one field and it is six figures and a three-month queue. The system becomes something you paid for and cannot move — that is what being locked in really means.
So what we are saying is not "we are cheaper", it isWe write the common failure points into the contract wherever they can be written。
Want to know roughly what a spec like yours costs? Get a range, with nothing to decide yet.
Free range estimateFirst, something people who sell systems will not say
Often you should not change systems at all.
After the diagnostic it often turns out the problem is not the system, it isYour process has never been written down in full. Swapping in a new system then just moves the same mess onto new screens — money spent, and six months later it still jams.
Often it is the process that needs rebuilding, not the software. So when we finish talking, the conclusion lands on one of these three —And on two of the three, we earn less.
01
Downgrade + bolt-on
Keep the existing system, drop to the cheapest tier, and patch automation in by API only where it genuinely jams. The smallest and fastest move, saving money first.
02
Freeze the version and take it over
The version is frozen and the whole thing comes to us; data and process untouched. Most of the industry will not touch someone else's code — we will.
03
Process rebuild + reconstruction
Only when the old one is genuinely beyond saving do we rebuild to your process. From NT$300,000, priced against your spec, with maintenance from NT$15,000 a month after launch.
A vendor who tells you to rebuild is not necessarily acting in bad faith — that may be the only tool he has. We do all three, so we have no reason to push you towards the most expensive one.
Which one are you? Thirty minutes of diagnostic tells you, and that step is free.
Book a free diagnosticHow implementation runs
Five steps, five safeguards
Every step exists to block one of the three failures above. The schedule and the amount come together, in writing, at step three.
Free 30-minute diagnostic
Guards against: being talked into it by a salesperson at the start
Johnny Yang talks to you, owner to owner. We do not send salespeople — the first call is with the person who will actually run the engagement.
Process mapping
Guards against: moving the old mess onto new screens
Write down the judgement in your head: how quotes are worked out, how delivery dates get agreed, who can give a discount, how exceptions are handled. Get this wrong and no number of systems will help.
Written feature list + fixed price
Guards against: budget creep halfway through
What will be built, what will not, how many stages, and what each stage delivers — in black and white. The amount and the schedule come together, not sign-first-add-later.
Staged development and acceptance
Guards against: seeing a screen for the first time six months in
Every stage ends with something you can actually click and use. Anything wrong gets fixed in that stage rather than discovered at the end.
We stay with you after launch
Guards against: being abandoned after acceptance
During the maintenance period, changing UI/UX, fields or report presentation for anything in the delivered written list costs nothing extra. Six figures to change one field is not something we do.
The first step is a free 30 minutes, and the person you talk to is the person who runs it.
Book a free diagnosticThe money
Said once, with nothing held back.
Few vendors publish prices, and fewer still publish what triggers an extra charge. These three tables are all of it — no fourth one appears when we talk.
① Three ways to take this on, and where the price differs
| How we take it on | Price | Who it suits | What we actually do |
|---|---|---|---|
| Downgrade + bolt-on | Quoted against the existing system and integration scope, usually below a rebuild | The current system still works, and only one or two steps jam | Map the process → drop to the cheapest tier → patch automation in by API |
| Freeze the version and take it over | Quoted after an assessment, then a monthly maintenance fee | The system itself works, but the original vendor stopped maintaining it, or you cannot change it | Freeze the version → decouple code and accounts → we take over maintenance |
| Process rebuild + reconstruction | From NT$300,000, priced against your spec | The process is definitely changing and the old system is beyond saving | On-site process mapping → written feature list → staged development and acceptance |
There is no "duration" column because the schedule depends on your spec and how fast requirements get confirmed. We give you that in writing along with the amount at the quoting stage, rather than committing to a nice-sounding number of days on a web page.
② Cost, and what is and is not included
| Item | Amount | When it is charged |
|---|---|---|
| Implementation (everything in the written feature list) | From NT$300,000, priced against your spec | In stages, tied to acceptance |
| Maintenance after launch | From NT$15,000 a month | Per month |
| Headcount | No headcount surcharge up to 200 people; above 200 quoted separately | — |
| Adjustments inside the written list (UI/UX, fields, report presentation) | No extra charge during maintenance | — |
| New features or modules outside the list | Quoted separately | Nothing starts until it is agreed |
| Third-party services (hosting, SMS, payments, maps…) | Charged by each provider, not through us | your own invoice |
Seen the costs and want to know where your spec lands?
Free range estimate③ What NT$300,000 actually buys
Each of these goes into the contract annex line by line. Be careful with anyone who wants a deposit before you have seen the list.
④ Whether it is worth it — count it yourself
We do not fill these three boxes in for you — once you have, you will know whether to do it, and that beats anything we could say.
Roughly how many months to pay back = ③ ÷ ①
This is an estimate for your own eyes, not a promise from us — only you know what goes in ①, we will not fill it in, and we do not guarantee the number you arrive at.
Filled in ① and the number looks too big? Then let us give you a range first, and come back and count again.
Free range estimateThe difference between the two approaches
It is not "packaged vs custom",
it is "an approach that fails vs one that does not".
To be fair: if your process looks like a generic form, a subscription is cheap and fast, and you should not come to us.Rent grows as you grow; an asset does not — that is the only difference.
The problem you will have in two years is being decided now
The system you build now,
Will it need rebuilding again in two years?
Most systems are built first and then somehow connected to AI. That means spending the money and going through the pain a second time two years later. We do it the other way round.
built for AI from the start
The data structures and the process are designed from the start on the assumption that AI will read, write and run against them. AI is not bolted on afterwards.
We use it ourselves every day
Brightstream is an AI-native company — what we give clients is what we run internally every day. Not theory.
The code and the data are yours
Once payment is settled you take the whole thing: source over git, data as a standard SQL dump, no lock-in and no proprietary formats. Licences for third-party services sit with those providers, and you take over those accounts yourself.
"You can take it with you" means these things in practice:The domain moved to your own hosting, third-party services rebuilt under your own accounts, the code in your own repository, and the database dumped into your own instance. Only when all four are done are you actually the owner rather than a tenant — with Sylvarc we did exactly those four first, and only then discussed the rewrite.
Want to know whether your current system can carry what is coming? Let us take a look first.
Free range estimateReal results · across industries
Three named cases, two of them rescued after someone else failed
That is not a coincidence. Most of the industry will not touch a system someone else wrote — which is exactly why we get called.
Two ERPs failed; the third one they came to us
Basis: adoption is Chingliu's own count before and after (under 20% before); dropped orders is the client's reported approximation, not a third-party audit.
The system was built years ago and never launched; we got them control of it back first
Basis: 4 is the number of stores the system covers (Taipei, Zhubei, Taichung, Kaohsiung); 2 is the number of delivery stages set in the contract. Both are statements of fact, not performance figures.
Turning a doctor's expertise into a brand that runs itself
Basis: 20,000+ is cumulative students since launch in 2019; 80% is the measured auto-reply rate after the 2025 AI rebuild. Both supplied by the brand.
These three sets of numbers came out oftheir industry and their size. Your numbers will be different — how different depends on your process, whether your team will drop old habits, and whether anyone inside is pushing it. We show these three to prove it can be done, not so you expect the same set.
These processes — we havebuilt ourselves
Not templated modules. We choose people, not industries — the list below is processes we have built, not the only industries we take.
Do not see your industry? Growing a new system around your process is exactly what we are best at.
Want to hear how it went in your industry? Fill in the details and I will tell you directly.
Free range estimateFirst, who we turn down
These four we turn down.
Writing this down is not posturing, it saves us both time. If one of these is you, our conclusion after talking would be not to do it anyway.
✕ Pure price comparison
If you already have three quotes and just want a fourth to push the price down, we are not interested. We are not the cheapest, and we say so at the start.
✕ No internal owner
If nobody inside the company can make a decision about the process, the project will jam. That is not a technical problem, it is nobody being able to make the call.
✕ Starting to build before the process is settled
If the process is still being argued over and will change again, writing code now just sets the mess in stone. Map the process first, then we start.
✕ A budget under NT$300,000
Below that we cannot deliver the quality we promise. In that case we will tell you to use an off-the-shelf subscription first.
And the limits of what is promised
I promise these,
and say plainly what I do not promise.
The three on the right are where the industry usually stays vague. We write them out, because a vague promise turns into a dispute on acceptance day.
- ✓Once the written feature list is agreed the amount is fixed. No "sign first, add later"
- ✓Staged delivery, with something usable at the end of every stage
- ✓Adjustments inside the written list cost nothing during the maintenance period
- ✓Once payment is settled, the source and the data go with you in full
- ✓From the first call to handover, the person you talk to is Johnny Yang
- ✕No guaranteed launch date — the schedule is tied to your spec and how fast requirements get confirmed. Committing to one up front only sounds good
- ✕No guaranteed staff adoption — that depends on how you push it internally; all we can do is build a system that does not get in the way
- ✕No guaranteed change in revenue — nobody can guarantee that for your business, and be careful with anyone who does
Still here and not put off? Then we can probably talk.
Free range estimate
Seven years founding companies. I have been the owner and hit the potholes, and I lead every engagement across both advisory and engineering.We do not send salespeople — call or add us on LINE and Johnny Yang answers; the person on that first call is the person who runs the engagement. Owner to owner, it goes faster.
The seven things owners ask most
What does a custom ERP implementation actually cost, and why will nobody publish a price?＋
We write it down: from NT$300,000, priced against your spec, with maintenance from NT$15,000 a month after launch. The price moves because of scope — how many processes to cover, whether existing systems have to be integrated, whether mobile is included; the difference is large. So we settle the scope first, put it in a written feature list, and then quote a fixed amount. No sign-first-add-later.
Where is the line on "no extra charge"? Will other fees appear later under another name?＋
The boundary is the written feature list at delivery: for anything on that list, adjusting UI/UX, changing fields, changing how a report reads, or smoothing the flow costs nothing extra during the maintenance period. New features and new modules outside the list are quoted separately. We fix that line at the start and do not play word games — because if we did, you would not come back.
Will you talk me into things I do not actually need?＋
The opposite tends to happen. The diagnostic often concludes "do not rebuild yet" — drop the current plan a tier and patch the jammed steps with APIs, or simply take over maintenance. The real problem is usually not the system but that the process has never been written down in full; swapping systems then just moves the same mess onto new screens.
How long does it take? Will it stall halfway?＋
The schedule and the amount are both given in writing at the quoting stage, split into stages with acceptance, and every stage delivers something you can actually click and use. We do not publish "live in N weeks" on a web page, because that depends on your spec and how fast requirements get confirmed — committing to it up front only sounds good.
Compared with a monthly cloud system, why should I spend this?＋
The two are not in conflict: if your process looks like a generic form off the shelf, a subscription is cheap and fast and you should not come to us. The companies that do come are the ones whose process cannot be changed, whose reports cannot be produced, and whose bill grows with every hire. Rent grows as you grow; an asset does not. That is the only real difference — the rest is sales talk.
Does the price go up as headcount grows?＋
No headcount surcharge up to 200 people; above 200 is quoted separately, and never sprung on you afterwards. The worst part of per-seat pricing is that the more you grow, the bigger the bill, so hiring becomes a punishment. We do not charge that way.
What if I no longer want to work with you?＋
The code and the data are yours: once payment is settled you take the whole thing — source delivered as a git repository, data as a standard SQL dump, nothing locked in a proprietary format. To be clear, licences for the third-party services the system uses (hosting, SMS, payments, maps) sit with those providers and do not transfer with our handover; you take over those accounts yourself. This is not just something we say — with Sylvarc, the first thing we did after taking over was hand control back to the client, and only then discuss the rewrite. A vendor who locks clients in does it because he knows he cannot keep them otherwise.
We are not the cheapest,
We putCommon failure points written into the contract。
We come back with the range a spec like yours usually lands in, and which parts would send the price up. That step is free and there is nothing to decide — you judge it yourself, and if it is not a fit, that is fine.
· Johnny Yang talks to you; we do not send salespeople
· A fixed price after the written feature list, no sign-first-add-later
· What it costs, what is and is not included, who I turn down, and what I do not promise — it is all above
Finally, to be clear:If the diagnostic says do not do it, that is what we tell you.We earn nothing that time, but you remember us when you really do need to move. No hard feelings if it is not a fit — we will not keep calling.
※ Amounts on this page are starting prices and starting monthly fees. The actual cost is quoted against the written feature list, not the numbers here.
※ Figures in the case cards were supplied by that client or measured over the period shown on the card. They reflect that client's industry and size and are not a guarantee of results for anyone else.
※ This service does not guarantee a launch date, staff adoption, or any change in revenue.