Insights · 11 min read
A practical guide to building a buy now, pay later platform: the core modules, realistic costs, and the compliance load that decides when the product ships.
By GGP Editorial
Buy now, pay later looks like a button at checkout. Behind that button is a lending business with a payment rail attached, and the gap between those two things is where most BNPL projects go wrong. Founders come to us with a checkout widget in mind and leave with a plan for a credit product, and the two are not the same budget.
I write this from the vendor side. GlobeSoft builds financial software, and since 2018 we have shipped trading systems, payment platforms, and accounting products for clients across the US, Brazil, South Africa, and Singapore. This guide covers what a BNPL platform actually has to contain, what it costs to build, and the regulatory load that decides whether and when the product ships. If you are still sizing the category, our breakdown of how much it costs to build a FinTech app is a useful starting point.
At checkout, the shopper picks "pay in 4" instead of a card. Three parties sit on the other side of that choice.
The merchant gets paid up front, in full, by the lender. The shopper receives the goods now and repays in installments. The platform sits in the middle, advancing the money to the merchant and collecting it back from the shopper over the repayment schedule.
The platform makes money three ways. A merchant fee, sometimes called a discount rate, of roughly 2 to 6 percent of each transaction, paid by the merchant for the lift in conversion and average order value. Late fees on missed installments. And interest on longer tenors, the six to thirty-six month installment loans that look less like "pay in 4" and more like a small consumer loan.
That last point matters for scope. "Pay in 4" is usually a short, interest-free product with no formal interest charge. A longer installment loan is a different product with a different risk profile and, in most markets, different rules. A platform that wants both is building two products, not one.
Here is the honest scope. Each row is a system you either build, buy, or rent from a partner, and each one has to work on day one.
| Module | What it does | Why it is not optional |
|---|---|---|
| Merchant integration | Checkout widget, APIs, and plugins for the big commerce platforms so merchants can offer BNPL | No merchants, no transactions, no product |
| Credit decisioning | Scores an application in seconds using identity, credit-bureau, and alternative data | The margin lives or dies on approval quality |
| Loan ledger and servicing | Tracks each loan, the repayment schedule, and the state of every installment | The ledger is the system of record and has to balance to the cent |
| Payments and settlement | Moves money to merchants and pulls repayments from shoppers | Money movement is the product |
| Consumer account | Lets shoppers see balances, make payments, and reschedule | Missed-payment self-service cuts defaults |
| Merchant and admin console | Onboarding, order lookup, refunds, and dispute handling | Operations runs here; without it every support question becomes an engineering ticket |
| KYC, AML, and fraud | Identity checks, sanctions screening, and transaction monitoring | Required before the first real customer |
| Compliance reporting | Regulatory filings and audit trails | Retrofitting it after launch is a rebuild |
Two of these deserve more detail because they are where BNPL projects burn the most time.
Credit decisioning. The product only works if you approve the right people fast. That means a rules-based first pass using identity, fraud signals, and a bureau pull, with a model behind it once you have repayment data. In the first year you have no data of your own, so you lean on bureau and alternative data and keep limits low. The engine also has to return a decision in a second or two. A checkout that makes a shopper wait is a checkout that loses the sale.
Disputes and refunds. Roughly 13 percent of BNPL transactions involve a return or a dispute, according to the CFPB's market work, so this is not an edge case. The platform has to handle a refund on a partly repaid loan, a merchant refund that arrives after the shopper has already paid in full, and a dispute where the merchant never shipped. Each of those has to end with the ledger balanced and the right party holding the right money.
BNPL is one of the more expensive FinTech products to build because of the number of systems and the compliance load. Based on current market ranges, a realistic picture looks like this.
| Scope | Typical range |
|---|---|
| MVP: pay-in-4, one market, one processor, rules-based underwriting | $70,000 to $150,000 |
| Full platform: multiple tenors, merchant API, risk models, collections | $200,000 to $350,000 |
| Multi-market or enterprise build: multi-currency, POS, local compliance per market | $350,000 to $700,000+ |
These are market ranges gathered from published estimates, not GlobeSoft quotes. The real number for your project comes from a scoped proposal, and the difference between a range and a quote is exactly what our how to estimate a custom software project guide explains.
Four things move the number more than anything else.
Number of markets. Every jurisdiction adds its own authorization path, its own data rules, its own reporting, and usually its own banking partner. One market is a product. Five markets is five compliance programs with one shared core.
Underwriting sophistication. A rules engine is cheap. A model that scores on transaction and cash-flow data is not. Most platforms start with rules and grow the model as repayment data accumulates.
Integrations. Every payment processor, bureau, and commerce platform is an integration with its own test environment and its own quirks. Five integrations cost more than five times one because you also build the layer that fails over between them.
Compliance scope. KYC, AML, sanctions screening, and reporting grow with each market and each product type. This is the slice that most often delays launch, because it involves people and approvals outside your team.
If you want the offshore cost picture specifically, our software development cost in China breakdown covers what a China-based team charges for this kind of work.
Regulation is moving fast in BNPL, and in different directions in different markets. What follows is accurate as of late 2026, but this is not legal advice, and you should confirm the current state with licensed counsel before committing.
United Kingdom. The FCA published its final BNPL rules in February 2026 as Policy Statement PS26/1, with the new regime taking effect in July 2026. Lenders need authorization, affordability checks, and clear financial promotions, and the sector's growth is the reason the rules came: the UK BNPL market went from about £0.06 billion in 2017 to over £13 billion in 2024, used by 10.9 million adults in the year to May 2024.
United States. The CFPB's 2024 interpretive rule that treated BNPL lenders like card issuers was withdrawn in May 2025, and the CFPB later said it does not intend to reissue it. That means the federal picture is lighter right now, but state lending and licensing laws still apply, and several states regulate BNPL directly. The US market remains the biggest, with the Federal Reserve estimating Afterpay at 34 percent, Affirm at 26, PayPal at 17, and Klarna at 15 percent of 2025 issuance.
European Union and beyond. BNPL generally falls under consumer credit rules in the EU, which means authorization and affordability duties in each member state. Other markets, from Australia to the GCC, are moving in the same direction: BNPL is being pulled into consumer credit law rather than left as a payment feature.
The practical consequence is that most new BNPL products do not hold their own lending license at launch. They run through a licensed lending partner or a banking-as-a-service provider, which gets them to market faster in exchange for a thinner margin and a real dependency on the partner. That is a legitimate way to start, and it is the same pattern we described for payments in how to build a cross-border payment platform.
You have three ways to get a BNPL product, and the right one depends on what you already own.
| Path | What you get | What it costs you |
|---|---|---|
| Build from scratch | Full control of the product and the margin | Longest time to market, full compliance burden |
| White-label a provider's platform | Launch in weeks, the provider carries the license | Thin margin, little differentiation, dependency |
| Buy composable infrastructure | Pick the pieces (decisioning, servicing) and assemble | Faster than a full build, still significant integration work |
If you are a retailer or a marketplace, white-labeling is often the sane choice. If you are a fintech and BNPL is the product, not a feature, then building is how you keep the margin and the differentiation, and the build is exactly the scope this article describes. Our FinTech MVP guide covers how to cut that scope down to a first release.
BNPL rewards starting narrow. Here is how we sequence it.
Phase one, the MVP. One market, pay-in-4 only, one payment processor, one or two commerce platforms, rules-based underwriting, a manual or semi-manual collections process, and a basic merchant and admin console. Expect four to seven months.
Phase two, the product. Add longer installment tenors with interest, a merchant API and self-serve onboarding, a first credit model trained on your repayment data, and automated collections. This is where the margin moves from thin to real.
Phase three, scale. Multi-currency, in-store POS, and one new market at a time, each with its own authorization, partner, and reporting. New markets are where most teams underestimate the work, so add them one at a time.
The thing to resist is building all of phase two and three under a phase one label. A BNPL "MVP" that includes every tenor, every market, and a full risk platform is not an MVP. It is the full product wearing a smaller name.
An MVP runs roughly $70,000 to $150,000, a full single-market platform $200,000 to $350,000, and a multi-market build $350,000 to $700,000 or more. These are market ranges, and the number for your project depends on markets, tenors, and integrations.
A focused single-market MVP takes about four to seven months. A full platform with multiple tenors and collections runs longer, and the authorization and partner-onboarding track runs in parallel and can outlast the build, which is why it starts first.
Usually yes, unless you launch through a licensed lending partner or a banking-as-a-service provider, which is how most new products start. Whether you need your own authorization depends on the market and how the product is structured, and it is a legal question to settle with counsel early.
If BNPL is a feature for your retail or marketplace business, white-labeling is often the faster, lower-risk choice. If BNPL is the product and the margin is the point, building is how you keep both, and it is a build best started as a narrow MVP.
Disputes and refunds, because a return on a partly repaid loan has to end with the ledger balanced and the right party holding the money. Roughly 13 percent of BNPL transactions involve a return or dispute, so it is a core flow, not an edge case.
GlobeSoft is a software development company founded in 2018, with 40-plus engineers and more than 300 delivered projects for over 100 clients across the US, Brazil, South Africa, and Singapore. We build custom software, mobile apps, SaaS platforms, FinTech and payment systems, ERP and CRM, and AI products on a stack that includes Java, Spring Boot, Go, Python, React, Vue, and Node. Our FinTech work includes a multi-market trading system spanning Hong Kong, US, and A-share markets, which you can read about in our multi-market trading system case study.
If you are planning a BNPL product, or a lending or payments product more broadly, and want to work through the scope, the compliance load, and a realistic budget before you commit, talk to us about your project.
Tell us what you are building and where you are today. We typically reply within 24 hours.
