Build on a platform or build custom?

The decision

An agency has quoted you ₹18 lakh for a custom build, and the thing they described looks a lot like the demo video of a platform that costs $79 a month. The question surfaces at scoping, and then again every time you need a new capability. It's high-stakes in both directions: custom costs 5–20x more upfront than standing on a platform, but the wrong platform costs a migration later — and a migration mid-growth is paid in months, not just money. The trick is knowing which of those two prices you are actually going to pay.

The questions that actually determine it

Where does your product differ from what the platform's demo shows?

Platforms are cheap where you're standard and brutal where you're not. Write down every place your product needs to behave differently from the platform's demo — not where you'd like it to differ, where it must. Then check each difference against the platform's extension model (apps, plugins, APIs). Differences the platform can absorb are free. Differences it can't are the whole decision. Most founders discover this list is far shorter than their pitch deck implies.

Is the workflow your moat, or just your plumbing?

Your moat is the thing customers pick you for. Your plumbing is everything that merely has to work: payments, invoicing, email, user accounts, admin panels. Moat workflows can justify custom investment. Plumbing never does — a custom-built invoicing system is money spent making yourself indistinguishable from everyone else. If you can't say in one sentence which workflow is the moat, that's the real problem, and no build decision should be made before it's answered.

What does the platform cost at 10x your current size?

Platform pricing is rent, and rent that looks tiny at the demo stage. A real-business tier runs $30–500/month base, plus per-transaction fees of 0.5–3%. At ten transactions a day the transaction fee is noise. At a thousand, it's a co-founder's salary. Run the platform's own pricing page at ten times your projected first-year volume before deciding. Sometimes the math kills the platform; more often it confirms that the rent is still cheaper than the build. Either way, you want the number now, not on the invoice.

How would you leave?

Do the escape analysis before committing, not after. Three things decide whether lock-in is an inconvenience or a hostage situation: can you export your data in a usable form, do you own your URLs and domain, and do you hold the customer relationships directly (emails, payment mandates) or does the platform? If all three come with you, leaving is a project. If the platform holds them, leaving is a hostage negotiation, and every future pricing change happens on their terms.

Your options, with honest costs and risks

The platform build

Shopify, Salesforce, or a vertical SaaS as the product, configured and themed. Two-year cost: $30–500/month base plus 0.5–3% of transactions — for many early businesses, ₹1–8 lakh ($1,200–10,000) total over two years, dominated by the transaction line if volume arrives. Risks: the ceiling is real and you find it suddenly; pricing changes happen to you, not with you. The lock-in is data shape and checkout flow; escaping later costs ₹5–20 lakh ($6,000–25,000) for a typical early product, plus the disruption.

The glue build

Existing tools wired together — a no-code front end, an automation layer, a spreadsheet or lightweight database behind it. Two-year cost: the tool stack runs ₹15,000–80,000/month ($180–1,000), so ₹3.6–19 lakh ($4,300–24,000) over two years, plus setup labor. This is the fastest way to test a workflow with real users. Risks: fragility at volume, per-seat fees stacking across five tools, and debugging that nobody owns. The lock-in is diffuse — each tool is individually escapable, which is exactly why this is a good temporary answer and a bad permanent one.

The custom build on managed services

Your own code on rented infrastructure. Build cost: ₹8–30 lakh ($10,000–35,000) with Indian agencies, $40,000–150,000 US/EU. Run cost: ₹8,000–50,000/month ($100–600) at early scale. Two-year total for an India build: roughly ₹10–42 lakh ($12,500–50,000), before any second phase of work. Risks: you pay the full price before learning whether the differences you built for matter; every future change costs agency hours; and you now own security, uptime, and maintenance. The lock-in is to your own codebase and whoever understands it — softer than platform lock-in, but real.

The hybrid

Platform core plus custom edge: Shopify for the store with one custom app for the workflow that makes you different, or a vertical SaaS with a custom layer talking to its API. Often the honest answer. Cost: platform fees plus a scoped custom piece at ₹2–8 lakh ($2,500–10,000). Risk: the seam. You depend on the platform's API remaining stable and permissive, and integration bugs live in the gap where neither vendor takes responsibility. The escape cost is the platform's escape cost — the custom edge usually survives a migration.

What I'd recommend

Buy plumbing, build moats — never the reverse. I ran a SaaS company for five years, and the most expensive engineering months were ones spent custom-building things a platform did adequately.

The default: if your differences-from-the-demo list is empty or cosmetic, use the platform and stop reading. If the list has one or two entries and they sit in your moat, go hybrid — platform core, custom edge, and spend only there. Go fully custom only when a difference in your core workflow is proven with real users — meaning people have used the standard version and the specific gap is what's blocking them — not merely asserted in a deck.

The mandatory comparison is the two-year total-cost table, not the upfront quote. Put all four paths in columns; for each, add build cost, 24 months of fees and run cost at your realistic volume and at 10x, and the escape cost if you're wrong. The upfront quote makes custom look expensive and platforms look free. The two-year table tells the truth, in both directions.

And do the escape analysis — data, URLs, customer relationships — before signing up, while it costs nothing. After launch it's a migration project with a ₹5–20 lakh ($6,000–25,000) price tag attached.

When this doesn't apply

• Your product IS infrastructure. If you're selling an API, a developer tool, or the platform itself, there is nothing to buy — the build is the business, and this framework has nothing to offer you.
• A compliance regime dictates architecture. Healthcare data residency, payment-card scope, financial audit trails: when a regulator constrains where data lives and how it moves, the platform-versus-custom question is answered by the compliance analysis, not by cost math. Get that analysis first.
• Existing systems constrain the choice. If your product must integrate deeply with a legacy system a customer or partner already runs, that system's integration surface — not this framework — decides what's buildable.
• You already have a strong technical team in-house. The cost math shifts materially when custom work is salaried rather than billed by an agency. The moat-versus-plumbing rule still holds; the ratios don't.