The "80% fit" trap
Every mid-market CTO has lived this. A Salesforce partner (or SAP, or ServiceNow — same pattern) shows up, scopes your workflow, nods wisely, and comes back with a quote. It starts at €60K for the platform fit-out. By the time implementation adds the 20% your workflow actually needs — the regulated approval chain, the non-standard data model, the integration with your legacy ERP — you are at €300K and the go-live date is 14 months out.
The trap is real: the 80% fit is where the vendor makes money. The 20% gap is where your business actually lives. Everyone involved knows this and nobody says it out loud.
The 7-criterion decision matrix
Score each criterion 1–5. Total over 25 means custom is likely the right call; under 15 means buy. In between is where honest scoping matters.
- Compliance specificity. Do you live under a regulator who asks for artefacts the platform does not produce natively (Slovenian ZVOP-2, Polish RODO, sectoral financial supervision)?
- Data residency. Must data stay in a specific jurisdiction, region, or on-prem? Every major SaaS is US-headquartered and US-court-reachable regardless of hosting claims.
- Workflow uniqueness. Does your core workflow have 5+ steps that do not map to a standard CRM pipeline or case object?
- Integration depth. Are you integrating with 3+ legacy systems (on-prem ERP, industry-specific middleware, custom hardware)?
- TCO horizon. Will the system be in production 5+ years? Custom amortises better over long horizons; SaaS seat fees compound.
- Team comfort. Does your team include developers, or will you rely on external consultants for every change?
- Vendor risk tolerance. Can you survive a 40% price increase at renewal? Can you survive the vendor being acquired and sunsetting your edition?
When custom wins
Custom is almost always the right call when you stack three conditions: regulated industry, genuinely unique workflow, and a multi-year TCO horizon. That is the profile where platform licenses become pure tax and implementation contractors become a permanent line item.
The modern custom stack is not what it was in 2015. You are not spinning up a bespoke Rails app with a team of six. A single senior full-stack engineer plus a focused agency can ship production-grade software in 3–4 months on Svelte 5 + Hono + Drizzle + Postgres — typed end-to-end, EU-hostable, and yours to own forever.
When buying wins
Conversely, buy when all three of these hold:
- Your workflow is genuinely standard — a sales pipeline that looks like other sales pipelines, case management that looks like other case management.
- Your team will use the platform as-is, not bend it. Every bend is a cost centre that grows over time.
- Your integration surface is shallow — a handful of standard APIs, not a legacy ERP from 2004 and a custom PLC controller.
If all three are true, Salesforce or HubSpot or ServiceNow really is the right answer. Save the custom build for the parts of your stack where you have a genuine edge.
3-year total cost of ownership — a real comparison
A 50-user deployment, regulated mid-market firm (insurance brokerage, €20M revenue), measured over 36 months. Rounded to the nearest thousand for clarity.
- Salesforce Financial Services Cloud: €90K licenses + €180K implementation + €72K ongoing support = ~€342K
- Custom build on Svelte 5 + Hono: €35K build + €30K implementation wraparound + €72K retainer (or zero if maintained internally) = ~€137K
- Delta: ~€205K over 36 months, on the side of custom. Plus you own the code.
These numbers are not hypothetical — they are pulled from two live scoping exercises we did with CEE clients in 2026. The Salesforce side came from a big-four-partner quote. The custom side is what we are currently charging. Your numbers will vary; the ratio usually will not.
The stack we use
For every Path C project, the default stack is Svelte 5 (frontend + dashboard), Hono (API), Drizzle + Postgres (data), optional private LLM on Hetzner EU. The reason is pragmatic: small surface area, fully typed, fast, and maintainable by any competent full-stack developer. No lock-in to a framework the vendor will sunset.
Galor-os itself — the platform running this site and our internal admin dashboard — is built on exactly that stack. We eat our own cooking.
Real result: Kosovni Odvoz
A different shape of the same argument. Kosovni Odvoz — a Slovenian waste-collection business — had zero online presence and a phone-based intake process. No off-the-shelf CRM solved this. They needed a custom site with per-city programmatic SEO across 13 Slovenian cities, plus a bespoke lead-routing dashboard so the operator knew which city converted which lead, daily.
We built it as a custom SvelteKit site. Zero ongoing platform fees. Zero reliance on paid acquisition for local intent. 7,000+ paying customers acquired through the site since launch. A Salesforce build for this problem would have been both more expensive and a worse fit.
Where to start
If you have priced a big-vendor build and hated the quote — start with a 3-day AI Opportunity Audit (€900). We will give you an honest scope comparison before you commit a cent to construction. Money-back if we do not identify €3,000+ in annual savings.