Bubble to Real Code: When to Migrate (and When Not To) 2026

What "real code" actually means

"Real code" is shorthand for a custom build on a stack you own. In 2026 that usually looks like:

• Frontend: Next.js or React with TypeScript

• Backend: Node.js or Python, served on Vercel, Railway, or AWS

• Database: Postgres via Supabase, Neon, or RDS

• Auth: Clerk, Auth0, or Supabase Auth

• CI/CD: GitHub Actions, staged deploys, migrations you can roll back

Bubble is a full-stack platform. It hides the database, the server, the deployment, and the frontend behind a visual editor. That trade is the whole point. You lose control and gain speed.

Real code reverses that trade. You get control back and pay for it in build time, salary, and engineering discipline.

Neither is better. They fit different stages of the same business.

When to migrate off Bubble: the five real triggers

Ignore the Twitter arguments. These are the triggers we see in actual client conversations that justify migration.

1. Your CPU bill is doubling every quarter

Bubble prices on Workload Units. As your app scales, you hit a curve where each new user costs disproportionately more compute than the last one. This is the single most common trigger.

We ran a three-week optimisation on MyAskAI, one of the leading AI knowledge-assistant startups. Their Bubble platform was straining under load. High CPU consumption was driving costs up and slowing the APIs their product depends on. We tuned the three highest-impact frontend pages, rebuilt the API layer around intelligent caching, and reviewed over 500 code references. Costs dropped by half. The platform runs roughly twice as fast.

That's the outcome you can often get without migrating. If a proper optimisation pass doesn't move the numbers enough, migration is the next honest conversation.

2. You need engineering rigour a no-code tool can't fake

Some builds need proper CI/CD, database migrations you can review, secure IAM, penetration testing, and audit trails. Regulated sectors and enterprise sales cycles will demand this in writing.

We took two Lovable-built apps for Astrid, a Norwegian software business, and put them behind a proper GitHub Actions to Supabase migrations to Vercel CI/CD pipeline across TEST and PROD. Doing a migration-replay audit from zero surfaced five hidden migration bugs the no-code tool had quietly left behind. The apps were live before the pipeline. They just weren't safe to change.

Bubble has a similar ceiling. It can pass a penetration test (we've done it, for Penny Ledger's UK tax platform, which files direct to HMRC), but it can't give an engineering team the branching, rollback, and code review workflow they'll expect if the product becomes core infrastructure.

3. A specific integration is blocking revenue

Bubble talks to most APIs fine. It struggles when the integration needs middleware, retry logic, webhook orchestration, or a data layer that sits between your app and someone else's system.

MatchNice is the clean example. Enterprise nonprofits run on Blackbaud, FundraiseUp, and half a dozen other systems, none of which want to change. Standard no-code tools couldn't reach into critical donor systems cleanly. We built bespoke middleware on AWS Lambda that ran the integration in the background. That middleware won them their first $300M+ nonprofit client and moved $300M+ in donor commitments through the platform.

The middleware could sit alongside a Bubble app or replace it. The question is whether the rest of the product also needs the same rigour.

4. Your team is genuinely ready to own the code

The dirty secret of migration is that a lot of teams don't have anyone to hand the new codebase to. They migrate to Next.js, hire two contractors, and are back where they started in eighteen months, this time with higher bills.

If you have a full-time engineer (or the budget to hire one) plus a technical co-founder or CTO who will review pull requests, you're ready. If you don't, migration will make your problem worse.

We built Junto, a group-travel marketplace, from scratch on Supabase, Railway, NestJS, Stripe, SendGrid, and Klaviyo. That's three apps (traveller, host, admin), 17 email triggers, admin-override database triggers, full auth. It works because they have the engineering ownership to run it after we hand it over.

5. Bubble genuinely can't do the thing

There is a small list of things Bubble genuinely can't do well: native mobile with heavy hardware integration, real-time audio/video processing, complex 3D rendering, extremely low-latency responses at scale. If your product needs one of these as a core feature, you were probably never a Bubble candidate.

For Just Video Walls, a US MicroLED video-wall manufacturer, we built a React and PDFMonkey technical-drawing and proforma PDF engine alongside the Bubble configurator — the diagram rendering was optimised from 12 seconds to 7 seconds. Bubble handled the quoting workflow, dealer roles, and HubSpot sync perfectly. Real code handled the piece Bubble couldn't touch.

Hybrid is often the right answer. You don't have to migrate everything.

When NOT to migrate off Bubble

These are the arguments we push back on, hard.

"Bubble looks unprofessional"

Your customers don't know or care what platform you're on. Penny Ledger files UK tax returns to HMRC directly, from a Bubble app, and passed a full penetration test. Formula Bot's backend still runs on Bubble; we only migrated the marketing site to Framer for SEO and speed. Nobody has ever churned because a URL had "bubbleapps" in it before the custom domain went on.

"Investors want real code"

Some do. Most don't, if the numbers are good. Pockla closed a £1.6M seed round backed by the Sidemen on a Bubble MVP we built. The seed round funds the migration if the traction demands it. Investors care about growth curves, not stack choice.

"It'll be cheaper long-term"

Almost never in the first two years. A proper Next.js build costs £40,000 to £120,000 to deliver, plus £3,000 to £8,000 per month to maintain and improve. Bubble sits around £150 to £500 per month for most SMBs. You need real growth to make the maths work.

Migration pays back when Bubble's compute cost, or Bubble's constraints, are directly costing you revenue.

"We hired a developer who doesn't know Bubble"

Hire a Bubble developer. Or hire Goodspeed. Don't rebuild the product because of one hiring decision.

Bubble vs custom code: honest comparison

The overlap in the middle is where most agencies get it wrong. They push migration when the client is still in Bubble's sweet spot. Or they insist on keeping Bubble when a proper rebuild is overdue.

The migration playbook (if you're actually going to do it)

We've done this enough times to give you the shape.

Phase 1: Audit and stabilise (2-4 weeks). Fix the fires first. Sometimes this alone kills the migration case. We rebuilt Symbio's cloud partnership platform without a full migration — 7 systems stabilised, 0 major incidents. Their team runs it themselves now.

Phase 2: Migrate the marketing site (3-6 weeks). Framer or Next.js on the marketing site is a fast win with real SEO and speed gains. Formula Bot moved to Framer, kept Bubble on the backend, and got 75% faster load times with zero SEO regression. That's the safest migration you can do.

Phase 3: Build the new backend in parallel (12-20 weeks). Ship the new stack behind a feature flag. Move users cohort by cohort. Never do a big-bang migration on a live product.

Phase 4: Cutover and decommission (2-4 weeks). Run both systems side by side for at least two weeks. Watch the metrics. Then turn Bubble off.

Total cost for a proper migration of a live SaaS: £60,000 to £180,000, delivered over 4-7 months. Bigger platforms cost more. Cheaper quotes usually skip Phase 1 or 4, and you'll pay for it later.

What Goodspeed would do

We build in Bubble when the situation calls for it. We build in code when the situation calls for it. We've migrated between the two, both directions.

If you're staring at rising Workload Unit bills, an integration you can't get past, or a customer-support conversation you keep having, book a Discovery Sprint. Two to three weeks, fixed price, money-back if you're not happy with the output. You walk away with an honest assessment: stay on Bubble and optimise, hybrid, or full migration, with the numbers behind each option.

We won't tell you to migrate if you don't need to. We've told plenty of prospects to stay put.