“Should we just use Firebase, or do we need a proper backend?” A founder asked us that on a scoping call, and it is the right question to ask before a single screen is designed. Our answer at Eightinity Technologies: for a new app with no existing server, start on a managed backend such as Firebase or Supabase. Build a custom one only when you already have a web product, or when a specific requirement rules the managed option out.
The reason is not that managed backends are better engineering. It is that your first version exists to test whether people want the product, and every week spent on servers is a week not spent learning that.
When is a managed backend the right call?
A managed backend gives you authentication, a database, file storage, and push notification plumbing without running servers. Firebase and Supabase are the two names founders hear most.
It fits well when:
- The app is new and there is no existing API or database to reuse.
- The data is simple: users, profiles, a few collections of records.
- You need to ship in weeks and learn from real users before committing to anything.
- You do not yet have anyone whose job is to operate a server.
Most seed-stage consumer and SaaS companion apps meet all four. For them, a custom backend in version one is usually cost without learning.
When does a custom backend earn its cost?
A custom backend is a service your team writes, deploys, secures, and monitors. That is a real, ongoing job, and it is worth it in a few specific situations.
- You already have a web product. The web app has an API and a database that hold your business rules. Building a second data layer for mobile means two sources of truth. On SOAPNoteAI, the iOS app consumed the existing web services, and the hard part was login, not the data. We covered that in the SOAPNoteAI web-to-iOS case study.
- Your business logic is heavy. Pricing engines, scheduling rules, or anything that must behave identically across platforms belongs on a server you control.
- You need relational reporting. Firebase's Firestore stores documents, not tables. If your product lives on joins and aggregates, a relational database is the natural home. Supabase is built on Postgres, which is why it often suits those products.
- A compliance or data-residency requirement rules it out. Check the provider's own documentation and your counsel before assuming either way.
Notice what is not on the list: “we want to scale.” Almost nobody outgrows a managed backend in year one, and the ones who do can afford to migrate.
What does a managed backend actually cost?
Firebase is free to start and charges by usage after that. According to its official pricing page, the Spark plan carries no cost and the Blaze plan is pay-as-you-go, billed by consumption past each product's free quota. Supabase's pricing page lists a free tier whose projects pause after one week of inactivity and a paid Pro plan that starts at $25 a month. Confirm current numbers on both before you budget, because quotas and rates change.
Two details matter more than the headline price. First, a free tier is for building, not for launching. A paused project or a quota ceiling during your launch week is a bad surprise. Second, usage-based billing rewards you for thinking about how often the app reads and writes data. A screen that re-fetches a list on every scroll is a bill waiting to happen.
We put hosting and backend at roughly 5 to 10 percent of post-launch spend for most apps in what an app's first six months after launch cost. It is modest until the product is AI-heavy, media-heavy, or syncs a lot of state.
What is the real risk of a managed backend?
Lock-in, and it is worth taking seriously. Firestore's data model and security rules are specific to Firebase, so moving off it later means rewriting the data layer and migrating records. That is a real cost. It is also a cost you pay only if the product succeeds, and at that point you have revenue and usage data to design the next backend properly.
You can reduce the lock-in without giving up the speed:
- Keep a repository layer in the app so screens never call the backend SDK directly.
- Put business rules that matter, such as pricing and permissions, in one place you control.
- Write down your data model on paper before you build it, not after.
- Export your data on a schedule from day one, so leaving is never a hostage situation.
The strongest counter-argument is that migrating later is painful, so you should build it right once. We disagree. The pain of migrating a product that works is smaller than the pain of building a custom backend for a product nobody wanted.
How do you decide in one afternoon?
You do not need an architecture review to make this call. You need clear answers to five questions about the product, written down before development starts.
- Does a web app, API, or database already exist that the mobile app must share?
- Is the data mostly simple records, or does it depend on joins and reports?
- Is there logic that must behave identically on iPhone, Android, and web?
- Does any regulation or customer contract dictate where data lives?
- Is anyone on the team able to run and monitor a server after launch?
Two or more yeses on the first four point toward a custom backend. A no on the fifth is a reason to hesitate even then, because a server nobody watches is a liability. If everything comes back no, rent the backend, ship, and revisit the decision when real usage gives you numbers instead of guesses.
The trade-off we accept is speed now against flexibility later. For a first release, we will take speed almost every time, and we say so up front so nobody is surprised in year two.
What should you ask an agency before they choose your backend?
The backend decision tends to get made quietly, in the first week, by whoever is writing the first code. Ask for it in writing instead. Our questions before scoping a native iOS build include backend readiness for this reason. Four questions are worth asking:
- Which backend are you proposing, and what is the specific reason it fits our product?
- Who owns the accounts, billing, and admin access, and are they in our company's name?
- What would it take to move off this backend in a year?
- What does the monthly bill look like at ten times our launch usage?
An agency that cannot answer the third question has not thought about your exit. An agency that proposes a custom backend for a brand-new app with no existing server should be asked what requirement justifies it.
FAQ
Should a startup use Firebase or build a custom backend?
A startup with no existing server should usually start on a managed backend such as Firebase or Supabase, because it removes server work and speeds up the first release. Build a custom backend when you already have a web product with its own API, or when heavy business logic or compliance needs rule a managed service out.
Is Firebase good enough for a production app?
Yes, for many production apps. Firebase handles authentication, a database, storage, and notifications at real scale. It fits simple data well. It is a weaker fit when your product depends on relational reporting, complex joins, or strict control over where data lives, so check those needs before committing.
How hard is it to migrate off Firebase later?
Moderately hard. Firestore stores documents with Firebase-specific security rules, so leaving means rewriting the data layer and moving records. You can limit the damage by keeping a repository layer in the app, centralising business rules, and exporting data regularly from the first day.
Should I use Supabase instead of Firebase?
Choose Supabase when your data is relational and you want Postgres underneath. Choose Firebase when your data is simple documents and you want its tight mobile tooling. Both are reasonable for a first release. Pick on your data model, and confirm current pricing on each provider's official page.
Not sure which backend your first version needs? Bring your feature list to a free 30-min call and we'll tell you what to build and what to rent →
