Who maintains it after you build it
Every migration conversation reaches the same question, usually around minute twenty: who patches this in fourteen months, when the person who built it has moved on and the team is still two people. It is a fair question and it deserves numbers instead of reassurance. The maintenance bill for a small custom stack in 2026 is mostly published on calendars you can read a year ahead, and it is smaller than the one people picture.
What the hosting actually costs
Cloudflare's Workers Paid plan starts at a $5 monthly minimum for the account and includes 10 million requests and 30 million CPU milliseconds. After that you pay $0.30 per additional million requests and $0.02 per additional million CPU milliseconds. Requests to static assets are free and uncounted. Cloudflare's own worked example prices a Worker serving 100 million requests a month, averaging 7ms of CPU each, at $45.40.
Supabase Pro starts at $25 a month, and the first Micro compute instance is covered by the $10 of compute credits the plan includes. That gets you 100,000 monthly active users, 8 GB of disk per project, 250 GB of egress and daily backups kept for seven days. The overages are unit-priced and public: $0.00325 per extra monthly active user, $0.125 per extra GB of disk, $0.09 per extra GB of egress.
Add it up and a small production app lands near $30 a month. Webflow's Premium site plan is $25 a month billed yearly. The hosting line barely moves, so anyone selling a migration on hosting savings is selling you something. What moves is what the next step up costs. Above Premium, Webflow's published tier is the Team platform plan at $2,500 a month on an annual contract, which buys governance and localization that a two person team mostly does not need. Above Micro on Supabase, the next instance is $15.
The upgrade work is on a calendar
Most of the ongoing work on a custom stack is runtime and dependency upgrades, and that schedule is public. Node 24 leaves active LTS support on 20 October 2026 and keeps security patches until 30 April 2028. Node 22 gets security fixes until 30 April 2027. Node 20 stopped getting them on 30 April 2026. You can put those dates in a calendar today and know roughly what 2027 costs you.
For a small app that is half a day of work once or twice a year. Bump the runtime, run the suite, deploy. It only gets expensive when there is no test suite, and that is worth saying plainly because it is the actual risk. A custom stack with no tests is harder to maintain than Webflow, since at least Webflow's vendor tests Webflow.
The part people leave out of the comparison is that a platform has maintenance too. You just do not get to schedule it. When a vendor reprices a tier, deprecates an API version or changes a submission limit, the work lands in your week whether or not you had room for it, and you cannot read the date off a public support calendar six months in advance.
What it looked like at Stay World Class
Stay World Class ran on Webflow for the front end and Xano for the back end. We replaced it with Next.js, NestJS and Supabase. Lighthouse went from 55.91 to 91 out of 100 and LCP got 77% faster. The hardest work was one-time: a four stage ETL that carried legacy record IDs into PostgreSQL UUIDs so no customer lost their history in the move.
What maintenance looks like there now is dependency bumps and the occasional schema migration. Nobody is on call in a way they were not before. The difference shows up when a page is slow. Now someone profiles it and fixes it that afternoon. On the old stack the fix was a support ticket and a hope.
When staying on the platform is right
If your site is a brochure with a contact form and nobody on the team writes code, Webflow at $25 a month is the right answer and will still be the right answer in three years. Migrating would hand you a maintenance obligation in exchange for lifting a ceiling you are not touching. That is a bad trade and you should not take it.
The same holds when hiring is the constraint. Custom software needs one person who can read a stack trace at 11pm, whether that is an employee or a contractor you trust on retainer. If your two people are a founder and a designer with nobody behind them, the platform is doing real work on your behalf. Move when the ceiling costs you more than the obligation would, and not a quarter before.
