Off-the-shelf SaaS is phenomenal. You start in 10 minutes, pay monthly, integrations are ready, someone else worries about uptime. Eventually every maturing business hits a ceiling: the tool is "almost right", but those 2-3% of mismatch cost hours of manual work every day. Here are five signals it is time to consider a custom solution - and a simple way to work out the maths.
Signal 1 - Your team works around the tool every day
Classic scenario: the SaaS "almost" fits, but does not support one specific flow. The team creates a workaround - an Excel sheet with macros, manual copy-paste between two tools, "evening scripts" running on the accountant's laptop.
It all works - until it doesn't. These are technical debts - one staff change or one upstream update can break everything in a single day.
Test: if the answer to "how does this work" contains "Beata does it on Friday afternoons", you have a problem.
Signal 2 - You pay for 100% of features, you use 20%
Most SaaS tools price per user and include everything. In practice you use maybe five features but you pay for the other thirty. At 30 people × $20/month that is $7,000 per year for features nobody has touched.
A custom application usually has a higher upfront cost but a flat running cost - it does not scale with the headcount. For teams above 15-20 users the ROI inflection often tips toward owning the software.
Signal 3 - The API is exactly what you need, but it is missing
The SaaS has a webhook for creating objects but not for updating them. Or API docs from 2020 that nobody updates. Or rate limits so low that your data flow doesn't fit.
At this point you start inventing creative workarounds: cron with a headless browser, scraping your own panel, "scheduled exports" to CSV plus an import script. Every such layer increases the chance of a failure mode.
Signal 4 - Sensitive data, and you don't know where it lives
GDPR made this complicated. Many good SaaS tools store data on servers outside the EEA. For some industries (health, education, finance, public procurement) that disqualifies the tool - even if it fits functionally. Standard Contractual Clauses help, but they are not always enough.
A custom application can run on infrastructure you choose: a VPS in Poland, on-premise, a private cloud. Control over data location becomes a feature.
Signal 5 - Your process is specific - not just "your company"
SaaS tools handle the typical process for an industry. If your flow is genuinely different - because it combines B2B and B2C, because your SLAs are unusual, because you margin per batch rather than per product - SaaS will fit only 60% of your work.
Forcing the fit (bending the process to the tool) usually erodes that competitive difference that lets the business exist. Better to build the tool around the process than the process around the tool.
How to work out whether it pays off
Simple arithmetic:
- Hidden SaaS cost: licences × 12 + (manual-work hours × hourly rate × 12).
- Custom MVP cost: project quote + 12 months of maintenance (typically 15-25% of project cost).
- Break-even: when (1) > (2) within 18-24 months.
If break-even lands inside two years, the custom MVP pays off. Most projects we deliver pay back in 12-18 months.
First steps
Don't start with code. Start with a scoping workshop - 90 minutes with the business owner of the process plus an engineer. You walk out with three things: a process map, the most pressing feature list, and a ballpark estimate.
See how we run custom software projects or book a free conversation - if SaaS keeps biting, in 30 minutes we'll tell you whether an MVP makes sense in your case.