SaaS vs custom: the real math (with concrete figures)

Every week, an SMB manager asks me the same question: "Should we go SaaS or custom?"

Wrong question. The right one: at what point does the total cost of a SaaS exceed that of a custom solution that does the same, better? And the answer, with figures, isn't the one most vendors want you to hear.

Why this question becomes critical in 2026

SaaS vendors raised prices by 30 to 80% between 2022 and 2026. Notion, HubSpot, Monday, Zendesk, Intercom — all moved their entry pricing from "affordable" to "significant". At the same time, the cost of custom development dropped thanks to AI and modern frameworks (Next.js, Supabase, n8n).

Result: the crossover curve has shifted. What was profitable in SaaS in 2022 isn't necessarily profitable in 2026.

The 3 real strengths of SaaS

Let's start with honesty: SaaS has real assets.

For market-standard functions (basic emailing, calendar, e-signature, micro-business accounting), SaaS is unbeatable. Don't go and build a Calendly clone — use Calendly.

The 3 weaknesses everyone discovers after 18 months

The problem is what SaaS demos never show.

Weakness 1: total dependence on the vendor

I lived this with three clients in 2024-2025:

• A vendor closed its "legacy" feature — the client lost 60% of their business logic.
• Another was acquired, the new owner tripled prices.
• A third changed its integration model, breaking 18 months of workflow.

You have no recourse. You don't have the code. You don't have the data in a portable format. You change tool, you start from zero.

Weakness 2: the hidden cost of business workarounds

As soon as your process leaves the standard covered by the SaaS, two things happen:

1. Either you adapt your process to the software (you pay in lost productivity).
2. Or you build workarounds (parallel spreadsheets, fragile no-code automations, team training).

In both cases, the real cost of the SaaS is no longer just the subscription. I saw a 40-person SMB pay €340/month for a CRM, plus 14 hours per week of an assistant maintaining manual workarounds. At €35/hour fully loaded, that's €24,500/year of "mask" — three times the tool's cost.

Weakness 3: cumulative cost

A SaaS at €200/month = €2,400/year. Over 5 years, €12,000. And that pricing will rise — SaaS vendors increase by 6-10% per year on average.

The real 5-year projection for a €200/month entry SaaS:

An equivalent custom solution (6-10 weeks of dev, €8,000 fixed-price delivery, €500/month optional maintenance): 8,000 + (500×60) = €38,000 over 5 years… but with no active maintenance = €8,000 + hosting (~€600/5 years) = €8,600.

The two cases aren't comparable without nuance. Let's see the real math.

The 5-year comparative calculation (3 scenarios)

I redid this calculation with my clients. Here are the 3 typical cases:

Scenario A — SMB using 80% of standard SaaS functions

Custom wins by 47% over 5 years. But careful — this is a specific case where human workarounds are high.

Scenario B — Hyper-growth startup

Near tie. Custom still has an advantage: you're the owner at the end.

Scenario C — Solo consultant / firm < 5 people

Very close. Here, custom is only profitable if you value code ownership and flexibility.

My 3-question decision framework

When a prospect asks me, I answer with three questions that clarify everything:

1. Does the market standard cover 80% of your need?

• Yes → stay SaaS. Don't pay to rebuild what Stripe, Calendly, Notion already do well.
• No → custom direction.

2. Is your business process a competitive advantage?

• Yes → custom mandatory. If your process is your value, you can't rent it to a vendor who standardises it across all their clients.
• No → SaaS is enough, as long as it does the job.

3. Can you absorb a forced vendor change?

• Yes → SaaS acceptable.
• No → custom, or SaaS with documented exit strategy (regular data export, technical independence).

If you answer "no" to at least 2 out of 3, custom wins in the long run.

The real good 2026 strategy: hybrid

The SMBs that optimise best don't pick a side. They do:

• SaaS for everything standard and shared (emailing, micro-business accounting, e-signature, video conferencing).
• Custom for what's differentiating: their business CRM, their client portal, their production tool.
• Automations (n8n, Make) as glue between the two, so everything talks.

It's the combination that gives both the speed of SaaS and the ownership of custom, without falling into total dependence or the cost of all-custom.

When custom is clearly a bad idea

Let's be honest — I regularly turn down custom engagements. Cases where it's the wrong answer:

• You're under 5 people and your process isn't yet stabilised: you'll change your mind 3 times in 2 years, better stay flexible on SaaS.
• Your need is ultra-standard (micro-business invoicing, email marketing, calendar): Stripe / Mailjet / Calendly do that better than I do.
• Your team has no digital appetite: a custom tool no one uses is double waste.
• Your goal is to save €2000/year on a SaaS: dev never amortises at that scale.

What you really gain with custom (beyond price)

Price isn't the first argument. What matters over 5 years:

1. You're the owner. Code, data, logic — all yours. You don't negotiate over your own tool.
2. You're aligned with your business. The tool wraps around your specifics, instead of forcing you to abandon them.
3. You control your roadmap. A feature you want? It ships when you decide, not when the vendor decides.
4. You depend on no one. If I disappear, another developer picks up in 2 weeks. You have the code.

Where to start

If you're hesitating, the healthiest move is a 30-minute audit. We list your current tools, cost out the real cost (subscription + workarounds), and I tell you frankly where SaaS is profitable and where custom would be smarter.

Book a slot — it's free, no commitment, and you leave with a clear costing. And if you want to see my price ranges for custom, they're public here.

To go further

• Related service: Custom application development
• Related articles: When to go custom · n8n vs Zapier vs Make