In short: the 80% rule. If an off-the-shelf package covers at least 80% of your process without workarounds, choose the package and integrate what is missing. Below that, or if your competitive edge lives in that process, custom software is usually cheaper and better over 5 years. In 2026 the tipping point shifts, because AI-assisted development makes custom work considerably cheaper.
Signs you have outgrown off-the-shelf software
Off-the-shelf software is built for the average customer. As long as you resemble that average, it is the best choice: live quickly, proven, and someone else pays for the roadmap. It starts to chafe the moment your process deviates. If you recognize 3 or more of these signs, you have probably outgrown your package:
- The "real" schedule, price list or inventory lives in Excel or Google Sheets next to the package.
- The same order, customer or invoice is entered in 2 or 3 systems.
- There are workarounds only 1 colleague understands, and that colleague can never be off sick.
- You pay for 40 modules and use 6.
- Customers ask for something the package cannot do: a portal, a status update, a quote with your own pricing rules.
- Every new employee costs €40 to €150 a month in licenses, even if they do 1 task in it.
- The vendor sets your roadmap, and your request has been "under consideration" for 2 years.
- Reporting takes days every month because the data sits in 3 systems.
The 80% rule in practice
Write out your core process in 15 to 25 steps, from first customer contact to invoice. Next to each step, note whether the package you are considering (or already have) supports it without a workaround. Below 80% coverage the question is not whether to build, but what. Above 80% you choose the package and integrate the rest. Then ask 1 extra question: does your competitive advantage live in this process? Bookkeeping is never a differentiator; the way you calculate quotes, plan routes or onboard customers often is. That is the part you want to control yourself.
5-year cost comparison
Off-the-shelf looks cheap because you pay monthly. Custom looks expensive because you pay up front. Put them side by side over 5 years, with a realistic SME scenario of 25 users:
| Cost item | Off-the-shelf (SaaS, 25 users) | Custom |
|---|---|---|
| One-off: implementation or build | €5,000 to €15,000 | €24,500 to €60,000 |
| Licenses per year | €22,500 (€75 per user per month), rising 5 to 8% a year | €0 |
| Hosting and operations per year | included | €1,200 to €3,000 |
| Maintenance and further development per year | included, on the vendor's roadmap | €4,000 to €10,000 (15 to 20% of the build cost) |
| Integrations | €2,000 to €8,000, often plus a monthly fee per integration | part of the build |
| Total after 5 years | €135,000 to €150,000 | €50,000 to €125,000 |
| What you own when you leave | an export, nothing more | your code, your data |
The ranges are deliberately wide: a simple SaaS tool at €25 per user tips the table toward off-the-shelf, an industry package at €150 per user tips it toward custom. The point is that you do the math with your numbers instead of looking only at the first invoice. For the build costs themselves, read what building a web application costs.
Hidden costs of workarounds and double entry
The biggest item appears in no quote. Do the math: 5 employees who each spend 30 minutes a day on double entry or checking is 2.5 hours a day, more than 550 hours a year. At €45 in labor cost per hour that is €25,000 a year, every year. Add the errors (wrong price, forgotten line, invoice 3 weeks late) and the risk that the one colleague who understands the workaround leaves. Over 5 years you are looking at more than €125,000 in hidden costs, more than enough to build and maintain a custom tool.
Hybrid: off-the-shelf package plus a custom integration layer
For most SMEs the smartest choice is neither, but both. The rule of thumb: buy what is commodity, build what is distinctive, and connect the two.
- Buy: bookkeeping (Exact Online, AFAS, Moneybird), payroll, email, payments, document storage. This is regulated, generic and better every year.
- Build: the workflow that only works this way at your company. Scheduling, quoting, a customer portal, a field-service app, a dashboard that brings 3 sources together.
- Connect: through the packages' APIs, so an order exists only once.
This works in regulated sectors too. A financial services firm keeps its core administration in a certified package and builds only the onboarding and the client portal itself. To see what such an integration looks like in practice, read connecting Exact Online to your online store; the approach to integrations and automation is the same for internal software.
How AI-assisted development makes custom software affordable in 2026
According to the Stack Overflow Developer Survey 2025, 84% of developers use or plan to use AI tools in their work, and 47% use them daily. That is no longer hype but a shift in what an hour of development time produces. In our practice, a senior maker working with Claude Code writes the predictable parts (data models, forms, tests, migrations, documentation) in a fraction of the time, and the time saved goes to what really matters: the domain model, security and architecture. An internal tool that took 12 weeks in 2023 now takes 6 to 8.
Two caveats. First, the tipping point of the 80% rule moves, but it does not disappear: you still do not build your own accounting package. Second, "generated quickly" is not the same as "built well". Software without tests, reviews and documentation becomes the new Excel: it works until it does not. Ask any studio that builds with AI about their test coverage, code reviews and security process.
RABZI indication: custom software at RABZI starts from €12,500 excl. VAT for an internal tool, from €24,500 for a process suite with integrations, and from €49,500 for a platform. If you are still torn between buying and building, a 1-week Discovery Sprint (€2,950, credited against a project started within 60 days) is the cheapest way to fill in the 80% rule with real numbers. All rates are on the pricing page.
Risks and how to cover them
Custom software carries 3 real risks, and all 3 can be managed with a good contract and a sensible stack.
Ownership of the code. Under Dutch law, copyright stays with the creator unless it is transferred in writing (article 2 of the Copyright Act). Put in the agreement that intellectual property transfers to you on delivery, or that you receive an unlimited, transferable license. Keep the source code in a Git repository in your name, not the vendor's.
Documentation and dependency. Demand a readable README, an architecture sketch, installation instructions and automated tests. The test: can another developer run the project locally within 2 days and change something safely? Also choose a mainstream stack (for example Node.js, PostgreSQL and React or Next.js) rather than an exotic framework, so there is always someone to be found.
Exit and continuity. Hosting and domains in your name, a data export in an open format, and a maintenance agreement you can cancel monthly. Escrow is unnecessary if you already have the code. Arrange security from day 1 (OWASP guidelines, updates, backups); the minimum you need for that is covered in NIS2 and the Dutch Cyber Security Act.
Worked example: a scheduling tool for an installation company
Take an installation company with 35 field technicians. Administration runs in AFAS, but the schedule lives in a spreadsheet the planner maintains 2 days a week. Technicians call in for their addresses, job sheets are filled in on paper and retyped in the evening, and invoices go out on average 3 weeks after the job.
The 80% rule: the package covers administration, purchasing and projects fine, but not the daily schedule or the job sheet. That is 25% of the process, and exactly the part where the margin sits. The hybrid solution: a custom planning board for the planner, a simple technician app for job sheets with photos and a signature, and an integration that pushes hours and materials straight into AFAS. Investment around €40,000 to €50,000, lead time 10 to 12 weeks. What it delivers: the planner spends 1 day a week instead of 2, the job sheet arrives the same day, and the invoice follows within a week. The shorter payment cycle alone pays for the build in the first year. A dashboard that brings schedule, hours and margin per job together is then a small extra step.
Frequently asked questions
What does custom software cost for an SME in 2026?
An internal tool covering 1 core process starts at RABZI from €12,500 excl. VAT, a process suite with multiple roles and integrations from €24,500, and a platform from €49,500. On top of that, budget 15 to 20% of the build cost per year for maintenance and further development, and €100 to €250 a month for hosting. Fixed price per phase, agreed in advance.
How long does it take to have custom software built?
A first working version is ready in 6 to 8 weeks, a full process suite in 10 to 16 weeks. The biggest delay is not the build but decision-making: who is allowed to do what, which exceptions count. That is why we start with a 1-week Discovery Sprint, so the scope is fixed before any code is written.
Who owns the code after delivery?
You do, if the contract says so. Without a written transfer, copyright stays with the creator. RABZI transfers intellectual property on delivery and places the source code in a repository in your name, including documentation and tests. Check this with every vendor before you sign.
What if the vendor goes out of business?
Your software keeps running, provided you have arranged 3 things: the code and documentation under your own control, hosting in your name and a mainstream stack. Another developer can then step in within days. So ask for the handover documentation at the quote stage, not when problems arise.
Conclusion
Choosing between custom and off-the-shelf is not a matter of belief but of arithmetic: coverage of your process, 5-year costs including the hidden hours, and the question of where your competitive edge lives. For most SMEs the answer is hybrid: an off-the-shelf package for the commodity, custom software for the process that makes you unique, and an integration in between. Want to fill in the 80% rule for your situation? Book a call: in 30 minutes you will know whether building, buying or connecting is the logical route.