Start with who changes the system.
A publishing platform succeeds when the content team can make normal changes safely. Ask who creates a page, who reviews it, which fields repeat and how often the structure changes. A custom stack is unnecessary if editors only need well designed content types. A visual builder is risky if every change alters business logic.
Map integrations before choosing the surface.
A form that emails one inbox is different from a customer journey that writes to a CRM, creates an account, checks inventory or triggers a staff workflow. List the data, owner, failure state and recovery path for each connection. Then decide whether platform integrations are sufficient or the workflow deserves application code.
Model the exit cost.
Confirm who owns the workspace, domain, assets and source. Ask how content exports, how redirects are managed and what happens when the supplier relationship ends. No platform removes lock in completely. The goal is a known and acceptable exit path.
Do not turn a website choice into an identity.
A capable studio can recommend a platform it does not sell when that is the right answer. DLVX uses Webflow, Framer and Next.js because the jobs differ. The decision belongs in discovery and should be explained in business terms before production starts.