Custom system or off-the-shelf tool? How to decide without guessing
An off-the-shelf tool (SaaS, software you rent by subscription) is enough when your company works the way most companies in your industry do, your team isn’t growing fast, and the tool connects to your other software without anyone retyping data. A custom system starts to make sense when people glue several programs together by hand every day, per-seat licences grow with the team, or the process you make money on differs from your competitors’. In many companies the best answer sits in between: SaaS where the process is standard, and your own panel and integrations where it isn’t.
In short
- An off-the-shelf tool is enough when your process matches the industry, the team is small and stable, and the integrations you need already exist.
- A custom system starts to pay off when per-seat licences and hours of manual retyping between programs keep growing.
- Compare costs over three years. In the example in this article SaaS is cheaper after year one, and the custom system comes out cheaper in year three.
- A hybrid often wins: SaaS where the process is standard, your own panel and integrations where your company is different.
- Before buying SaaS, check data export and the API; in a contract for a custom system, secure access to the code, database and hosting.
I’m writing this from both sides. At Reefa, my B2B cleaning company in Kraków and Katowice, I built the panel that runs the whole business. We still issue invoices in an off-the-shelf app, because building our own would add nothing. Below are the five questions I start every conversation with, and a worked three-year cost example.
Five questions that settle it
Each question tips the scale one way. Answer them with the people who actually do the work, not just their manager. They’re the ones who know where the workarounds are.
1. How different is your process from the rest of your industry?
Off-the-shelf tools are built for the average company. If you sell, invoice and run HR like everyone else, SaaS will do it well and for less. Trouble starts where your company does something the tool never anticipated: its own way of pricing jobs, paying crews across many sites, quality checks on random samples, or one process running across several legal entities.
A quick test: list the steps of the process and mark the ones the tool handles without workarounds. If the workarounds (extra fields, a spreadsheet on the side, “we just remember to do it”) sit on the steps that earn you money, that points towards your own system.
2. How many tools do people glue together by hand?
Count the programs between which someone retypes data: the CRM, the schedule spreadsheet, the invoicing app, the bank, the WhatsApp group with the crew, the inbox. Every handover costs time and invites mistakes. If there are one or two such places, an integration between existing tools is often enough. If there are five or more, and someone in the office spends most of the day moving data around, one place where all that data meets usually works out cheaper.
3. Who owns the data, and what happens when you switch vendors?
Ask the SaaS vendor three things. Can you export all your data in a readable format? Is there an API (the interface other software uses to read and write data)? What does leaving cost? With a custom system the data sits in your own database, but that has to be in the contract too: database access, source code, documentation and hosting accounts in your company’s name. No export in SaaS and no such clauses with a developer are the same problem. Either way, you depend on one company.
4. What will per-seat licences cost over three years?
A per-user price looks harmless at five people. At fifty, including field staff who need two buttons, it can become the biggest line in the IT budget. Always count three years and include the growth you’re planning. Check whether pricing depends on tiers as well: the one feature you need is often only in the more expensive plan, paid for every user.
5. What does the system need to connect to?
List everything the system has to exchange data with: accounting, the bank, Poland’s KSeF e-invoicing through your invoicing app, calendars, phones, machines on the shop floor, a client portal. A SaaS product usually has a handful of ready integrations and plenty of gaps. If your key connections are on the vendor’s list, the choice is easy. If they have to be built, you end up writing custom software anyway, just around someone else’s system.
Reading your answers
| Question | Leans off-the-shelf | Leans custom |
|---|---|---|
| Process | same as the industry | you make money on how you differ |
| Manual gluing | 1–2 places | 5 or more |
| Data | export and API available, leaving is cheap | no export, or leaving is expensive |
| Licences over 3 years | small, stable team | growing team, many field staff |
| Integrations | ready at the vendor | have to be built |
Three answers in the right-hand column are a good reason to cost out both options. One isn’t.
What each option costs over three years: an example
The figures below are a made-up example. They show how to do the maths, not the price of any real product or my own rates. Plug in your own numbers.
Assumptions for an example service company with an office and field crews:
- 30 people use the system in year one, 40 in year two and 50 in year three, 10 of them in the office
- the SaaS costs PLN 150 per user per month excluding VAT, plus PLN 15,000 to set up
- with SaaS the office still retypes data for 10 hours a week, with a custom system for 2 hours (50 weeks a year at PLN 50 an hour)
- custom system: PLN 120,000 to build, PLN 3,000 a month for maintenance and development
- hybrid: SaaS for the 10 office staff only, plus a custom panel, integrations and a crew bot for PLN 70,000, with PLN 1,500 a month to maintain
| Cost over 3 years, excl. VAT (example) | Off-the-shelf | Custom | Hybrid |
|---|---|---|---|
| Setup or build | PLN 15,000 | PLN 120,000 | PLN 85,000 |
| Licences | PLN 216,000 | PLN 0 | PLN 54,000 |
| Maintenance and development | included | PLN 108,000 | PLN 54,000 |
| Manual retyping | PLN 75,000 | PLN 15,000 | PLN 15,000 |
| Total | PLN 306,000 | PLN 243,000 | PLN 208,000 |
The timing matters more than the total:
| Cumulative cost (example) | Off-the-shelf | Custom | Hybrid |
|---|---|---|---|
| After year 1 | PLN 94,000 | PLN 161,000 | PLN 126,000 |
| After year 2 | PLN 191,000 | PLN 202,000 | PLN 167,000 |
| After year 3 | PLN 306,000 | PLN 243,000 | PLN 208,000 |
After year one the SaaS is PLN 67,000 cheaper. After year two the gap is down to PLN 11,000, and during year three the custom system pulls ahead. That’s why a one-year comparison almost always points to SaaS, and with a growing team it’s misleading.
Now change two assumptions: the team stays at 10 people and the process is standard, so nobody retypes anything. SaaS then costs PLN 69,000 over three years (PLN 15,000 setup plus 10 seats at PLN 150 for 36 months) and beats any custom system. There’s no single answer here. There are two variables that decide almost everything: how many people pay for a seat, and how many hours someone spends moving data by hand.
I’d add one risk the table doesn’t show. If the process isn’t written down, the build drags on and the budget grows. That’s why I always map the process and count the hours before I quote. You can see what manual work costs your company with the automation savings calculator.
The middle path: SaaS at the core, your own panel around it
A hybrid keeps the off-the-shelf tool where it does its job well and adds your own software only where your company is different. In practice that’s usually three things. Integrations move data without retyping. A simple interface lets field staff work without a licence each. A panel shows in one place what’s currently scattered across several programs.
Here’s how it looks at Reefa:
- invoices are issued in inFakt, a Polish invoicing app; every morning the panel pulls them from several accounts and matches them to sites by tax ID
- the bank is connected through PSD2 open banking, so the panel checks on its own that every transfer has an invoice or a cost document
- field staff have no app and no licence: they start and finish shifts, send photos and report missing supplies through a Telegram bot
- contracts are signed on our own signing page with an SMS or email code, and signatures made with Profil Zaufany, Poland’s free government e-signature, are verified automatically
I didn’t write my own invoicing software or my own messenger. Nobody would have paid me for that work. I built what no off-the-shelf product did the way Reefa works: schedules and shifts across many sites, finances from several accounts in one view, quality audits and reports for clients.
A hybrid has one condition: the SaaS must have an API and a full data export. Check this before you buy, not on the day you want to plug something into it.
When not to build your own system
I advise against a custom system when:
- the process is standard, like accounting, payroll, email or a simple sales CRM; off-the-shelf tools do these well, and the vendor updates them when the law changes
- only a few people use the system and that number won’t change
- you can’t yet describe how the process should work; then you start with a process map, not with code
- you need a report rather than a system; sometimes a well-built spreadsheet or a dashboard on data you already have is enough
- nobody in the company will own the system; custom software needs someone who spends a few hours a week deciding what matters, checking new versions and reporting bugs
- you expect software to fix a messy organisation; a system speeds up a process that already works, but it won’t fix a broken one
If several of these fit your company, pick an off-the-shelf tool and invest in rolling it out properly. That’s a good decision too, and often the best one.
How Reefa’s panel grew out of everyday questions
Reefa’s panel wasn’t built from one big plan. Every module started with a question I couldn’t answer quickly. Who was on site yesterday? Which client hasn’t paid? Does every bank transfer have a document? What should we send the client at the end of the month?
| When (2026) | Need | What got built |
|---|---|---|
| June | know who hasn’t paid | daily invoice sync from inFakt and a morning message in Telegram |
| July | all invoices in one view | several inFakt accounts in one system with a switcher between them |
| July | an orderly cash register | cash module: 312 entries from 21 months imported and reconciled to the grosz |
| July | overdue invoices | debt collection in 9 stages, from 2 days before the due date to 60 days after |
| August | certainty that every transfer has a document | monthly bank-to-document reconciliation across all accounts |
| October | client reports without manual work | a PDF visit report on the 5th of each month, switched on for all clients on 16 October |
Building this way has an upside: every module solves a real problem from day one, and nobody asks why they should use it. It also has a cost worth knowing up front. After a few months of adding modules, the panel menu had 33 sections. On 15 October 2026 I merged them into 13 sections with tabs and added worker, site and client cards. If you build in stages, plan a clean-up like that every so often.
The second lesson: the field team got a bot, not an app. In August 2026 alone it recorded 1,100 hours of work across 364 shifts, and by mid-October the system held about 2,700 logged shifts. Not every site uses the bot yet, so those numbers don’t cover all the company’s work. They do show that collecting data from the field doesn’t always need a separate app.
I describe the whole system in the Reefa operating system case study and the bot in the field-team bot case study.
How to make the call in your company
- Answer the five questions with the people who do the work.
- Count three years of licences, including planned team growth.
- Count the hours spent retyping data and multiply them by the hourly cost.
- Ask the SaaS vendor about data export, the API and the cost of leaving.
- Compare the three options in a table like the one above, using your own numbers.
If an off-the-shelf tool turns out to be enough, I can help you choose it and roll it out as part of IT rollouts and fractional CTO work. If it comes out as a custom system or a hybrid, I build custom business systems, from PLN 60,000 excluding VAT. If you’re not sure yet, start with the AI and process audit (from PLN 6,000 excluding VAT). Two weeks: interviews with your team, a process map, counted hours and a rollout plan with order and budget.