Build vs Buy: What Founders Get Wrong About Templates

Every week a founder DMs me the same question: "Should I build my React Native app from scratch or buy a template?" Here's the framework I use to answer it, plus the five wrong assumptions that make the debate louder than it needs to be. Differentiation lives in the product (AI pipeline, marketplace mechanics, workflow)? Buy a template and spend your engineering budget on the differentiation. Differentiation lives in the infrastructure (custom protocol, native audio pipeline, novel storage engine)? Build from scratch. Roughly 90% of consumer and prosumer apps live in bucket one and pretend to live in bucket two. "Build or buy" is really four separate questions wearing a trench coat. Answer them separately or the debate never resolves. "Build or buy" is a stand-in for four separate questions: Will my app look like shovelware? Will I get stuck when I need to change something? Is a $79 template recycled GitHub code? Am I "cheating" if I didn't write every line? Each has a different answer. Bundling them together is why the debate never resolves. Both paths start from the same primitives: React Native, Expo, a DB, an auth provider. A template just made 200 opinionated decisions for you: the same 200 you'd make identically after two months of research. # What a template saves you from doing yourself npx create-expo-app my-app # ... then 60 hours of: # - navigation library selection # - auth flow # - form validation # - RLS policies # - Stripe webhook handlers # - EAS build config # - push notification setup # - offline handling # - error boundaries # - image caching You still have full control. Templates ship source code. You just skip the mandatory chores. Take away the color, logo, and copy from the App Store's top 100. What's left? Tab bar, list, detail, modal, auth. Uniqueness comes from brand, motion, and product concept, not from the base template. A React Native template that uses NativeWind can be re-themed in an afternoon: // tailwind.config.js: the entire re-branding surface for a NativeWind template module.exports = { theme: { extend: { colors: { primary: "#FF5722", // your brand surface: "#0B0F14", accent: "#00D4AA", }, }, }, }; Reading a well-written template teaches you faster than writing one from scratch. You're editing production patterns on day one instead of inventing bad ones over three months. Full-stack templates like Applighter ship the entire vertical: Postgres schema + migrations RLS policies (see Supabase's RLS docs) Edge functions for server-side logic Storage buckets with signed URLs Auth (email, OAuth, Apple) Streaming AI responses Push notifications If a "template" doesn't ship all of that, it's a UI kit. Charge accordingly. Here's the difference in practice. A UI kit hands you this: // components/TodoList.tsx const TODOS = [ { id: "1", title: "Buy milk", done: false }, { id: "2", title: "Ship app", done: false }, ]; export function TodoList() { return <FlatList data={TODOS} renderItem={({ item }) => <Row {...item} />} />; } A full-stack template hands you this, plus the migration and the policy that make it safe: -- supabase/migrations/0001_todos.sql create table todos ( id uuid primary key default gen_random_uuid(), user_id uuid not null references auth.users on delete cascade, title text not null, done boolean not null default false, created_at timestamptz not null default now() ); alter table todos enable row level security; create policy "owner reads own todos" on todos for select using (auth.uid() = user_id); create policy "owner writes own todos" on todos for insert with check (auth.uid() = user_id); // hooks/useTodos.ts export function useTodos() { return useQuery({ queryKey: ["todos"], queryFn: async () => { const { data, error } = await supabase .from("todos") .select("*") .order("created_at", { ascending: false }); if (error) throw error; return data; }, }); } That gap (mock array vs. schema + RLS + typed client) is the entire difference between $29 and $200. The math: Approach Cost Time Full-stack template $79–$200 2–6 weeks Solo contractor build $6k–$12k 8–12 weeks Agency build (median) $18k+ 3–6 months DIY founder from scratch $0 cash / 3–6mo 3–6 months The template pays for itself the day you download it. Dimension Build from scratch Buy a template Time to first App Store submission 3–6 months 2–6 weeks Upfront cash $0–$30k $50–$200 Full source ownership Yes Yes (with legit templates) Time on "solved problems" 60–80% 5–10% Backend included No Yes (full-stack) You're building infrastructure, not a product. Your app's core primitive doesn't fit React Native (real-time sub-16ms, custom Metal shaders, on-device model runtime). You have engineers on payroll already and no launch pressure. Compliance requirements (SOC 2, FedRAMP, HIPAA) demand ownership of every line. For every one of these, we see 50 apps that fit none of them but choose to build from scratch anyway. That's a hobby, not a decision. App is a variant of a solved category (chat, notes, tracker, marketplace, AI wrapper) Differentiation is UX or product logic, not architecture You want to spend month one on your unique value, not on auth boilerplate Source code included and readable (not obfuscated, not a paid-SaaS wrapper) Backend included (not just UI screens) Commits in the last 90 days Refund policy in plain English The author uses their own product Almost every successful indie mobile app in the last three years followed this loop: Buy a full-stack template Ship branded MVP in 3–6 weeks Get real user data Rewrite the two or three modules that turned out to matter They didn't "buy" or "build." They bought the boring 80% and built the interesting 20%. Where is my differentiation, and what's the fastest path to it? Spend your engineering budget where it differentiates. Everything else is a chore, and chores have a market price now, usually under $200. Disclosure: I write templates for a living at Applighter (Expo + Supabase, full source, no vendor lock), so I'm not neutral here. What did you start your last app from, scratch or a template, and would you make the same call again? Drop it in the comments.