The real unit of comparison: exit cost
No-code tools are excellent at the start and everyone prices the entry. The honest comparison prices the exit: what it costs, in money and possibility, to leave the tool when the business outgrows it. Custom development's exit cost is low (you own the code; portability is the point). No-code's exit cost is the entire rebuild — because the logic lives in the platform.
What no-code genuinely does well
Validation speed. A no-code MVP answers "does anyone want this" in weeks for hundreds, not months for thousands. Marketing sites, internal tools with modest complexity, single-process automations — no-code is the right tool and the ROI is embarrassing for custom work at that scale. Anyone selling you custom code for a hypothesis you have not tested is selling you furniture for a house you have not visited.
Where it quietly breaks
- Logic ceilings — the day your product needs real application logic, workflow builders become the constraint they were marketed as freedom from.
- Cost scaling — per-record, per-seat, per-action pricing compounds with your success.
- Performance ceilings — you cannot engineer around the platform's rendering.
- The exit rebuild — every rule you built in a visual editor is re-implemented, by hand, from inspection.
The decision pattern that ages well
Validate no-code, operationalize custom. Test the hypothesis where it is cheap; when it survives, rebuild deliberately — data model first, then logic, then surface — while the no-code version keeps earning. The migration is a known shape (inventory → map → rebuild → cutover), the same discipline as any migration.
The uncomfortable middle truth
Most businesses never outgrow no-code, and that is fine — the trap is only for those who do and discover it during a fundraise, a due-diligence pass, or an outage. The audit that tells you which side you are on costs one conversation. The brief routes there.