Where Digital Lending Money Is Moving

There are two very different numbers people quote when they talk about the size of this market, and confusing them is the fastest way to lose credibility with an investor on page three of your plan.

The first is origination volume: the value of loans actually funded through digital channels. Mordor Intelligence puts the US digital lending market at roughly $303 billion for 2025, growing toward $560 billion by 2030 (Mordor Intelligence, 2025). That is the pool you are lending into.

The second is the digital lending platform software market: what banks, credit unions and non-bank lenders spend on the technology that originates and services those loans. Estimates for 2025 cluster between $12.6 billion and $15.6 billion globally, with forecast growth rates spread across a wide 14% to 28% CAGR band depending on how each firm draws the category boundary (Grand View Research; Fortune Business Insights). One widely circulated 2025 forecast projects the platform market reaching $39.8 billion by 2033 (Research and Markets, 2025).

Two markets, two orders of magnitude. Your plan should name which one you are addressing in the first paragraph of the market section.

What is actually driving growth

Growth in this category is not coming from consumers suddenly wanting loans. Demand for credit is broadly cyclical. What has changed is the cost and speed of originating a loan. Three shifts matter to your plan:

• Bank-partnership infrastructure matured. A founder no longer needs a charter or 50 state licences on day one. A sponsor bank originates, the platform distributes and services, and revenue is a fee on funded volume.
• Cash-flow underwriting went mainstream. Bank-transaction data supplements or replaces thin bureau files, which opened lending to near-prime and small-business segments that traditional scorecards declined.
• Embedded distribution beat standalone brands. The expensive part of lending is finding a borrower at the moment they need credit. Vertical software companies already own that moment, which is why embedded lending is where most new platform revenue is being created (Lendflow).

Category scale is not theoretical. Funding Circle alone has originated more than $20.2 billion to over 135,000 small businesses across 700 industries (Upwise Capital, 2025). Upstart operates as an AI underwriting layer supplying banks and credit unions rather than lending alone. Zopa started as a peer-to-peer marketplace in the UK and converted into a licensed bank in 2020. LendingClub followed a similar path, acquiring a bank charter after years as a marketplace.

That pattern is worth naming in your plan. The successful platforms nearly all migrated toward owning either the charter or the distribution. Sitting permanently in the middle, renting a bank and buying traffic, is the position with the thinnest defensibility. Investors know this. A plan that shows a credible path to one end or the other reads very differently from one that assumes the middle is a stable place to live.

UK and European context

The UK market is smaller but structurally cleaner. One FCA authorisation covers the whole country, so there is no equivalent of the 50-state licensing grind. That single fact changes the shape of a UK plan: less capital tied up in licensing, more of the launch budget available for the credit model and acquisition. The trade-off is that the FCA's Consumer Duty and CONC sourcebook set a higher documented standard for affordability assessment and forbearance than most US state regimes, and the regulator expects to see that evidenced before authorisation, not after.

What Founders Ask First

These come up in almost every scoping call we run for a lending client. They are worth resolving before you write the plan, because each one changes the financial model materially.

What is the difference between a digital lending platform and a loan origination system?

A loan origination system (LOS) is software: it captures an application, runs decisioning rules, generates documents and hands off to servicing. Vendors like nCino, Blend, LoanPro, Mambu and TurnKey Lender sell into that slot. A digital lending platform is a business that originates loans, and it may buy an LOS rather than build one. Confusing the two produces a plan that describes a software product while the financial model forecasts interest income, which no credit investor will accept.

Can a digital lending platform operate without holding a lending licence?

Yes, through a sponsor bank arrangement. The bank is the lender of record and holds the charter; the platform handles acquisition, application flow, underwriting support and servicing, and is paid a fee. This is the standard route to market for a new US platform because it removes the multi-year state licensing timeline from the critical path. It does not remove regulatory obligation: the bank will impose its own oversight programme on you, and the CFPB supervises nonbank lenders and service providers on disclosures, fair lending, complaint handling and data accuracy under TILA, ECOA and FCRA (InnReg).

What is the true lender rule and why should it be in my plan?

When a platform partners with a bank so the bank can export its home-state interest rate into states with lower usury caps, courts and state regulators may look past the paperwork and ask who really bears the economic risk of the loan. If the answer is the platform rather than the bank, the loan can be recharacterised and the rate cap of the borrower's state applied retroactively. That is an existential risk for a high-rate product. Plans that go to institutional investors should state which states are in scope, what rates are charged, and how much of the loan the bank retains.

How much capital do I need if I lend from my own balance sheet?

More than most founders expect, and it scales with volume rather than being a one-off. As a working rule, expect to hold 10% to 20% of outstanding balances as equity behind a warehouse facility, with the lender advancing the rest. On a $20 million book at a 15% advance-rate haircut, that is roughly $3 million of your own money permanently tied up and growing with the portfolio. This is the single biggest reason first-time founders choose the fee-based model.

What is a forward-flow agreement?

A committed arrangement where a credit fund or bank agrees in advance to purchase loans you originate, at defined credit criteria and pricing, up to a monthly or quarterly cap. It converts your business from balance-sheet lending into origination-for-fee, and it is the funding structure most early-stage platforms actually raise. A signed or term-sheeted forward-flow facility in the appendix of your plan does more for an equity raise than another ten pages of market analysis.

Launch Capital, Line by Line

A realistic launch budget for a US digital lending platform on the bank-partnership route runs $120,000 to $520,000 (roughly £95,000 to £410,000) before any loan capital. Third-party estimates put the range at $100,000 to $500,000, which corroborates that band (Businessplan-templates, 2025). Balance-sheet lending sits on top of that figure, not inside it.

The distribution of that spend surprises people. Most founders arrive assuming engineering is the dominant line. In a lending business it usually is not. Licensing, compliance and credit infrastructure typically consume more of the pre-revenue budget than the product build.

Cost breakdown

• Multi-state lending licences (US): $45K–$220K for 8 to 25 states. Filed through NMLS using the MU1 company form, with application fees typically $500 to $3,500 per state, surety bonds commonly $25K to $100K per state, minimum net worth tests, and a registered agent in each jurisdiction (CSBS).
• FCA consumer credit authorisation (UK): £12K–£45K covering the non-refundable application fee, a compliance consultant, the policy and procedure suite, affordability assessment framework and SM&CR setup.
• Loan origination and servicing software: $6K–$250K in year one. Small-lender SaaS runs $500 to $3,000 per month; mid-market licensing runs $30,000 to $120,000 annually before implementation; enterprise deployments can reach $250,000 or more in total first-year cost (Timvero, 2026).
• Credit and identity data vendors: $18K–$60K. Bureau pulls, bank-transaction aggregation, identity verification, device intelligence and fraud screening, most of which carry annual minimum commitments that a low-volume launch will not use up.
• Underwriting model and fair-lending testing: $25K–$90K. Scorecard or ML model development, disparate impact testing, and model documentation your bank partner's risk committee will demand before first origination.
• Regulatory counsel: $20K–$85K for the bank partnership agreement, a true-lender opinion where relevant, borrower terms, and TILA-compliant disclosure templates.
• Borrower experience build: $30K–$140K for application flow, decision screens, document upload, e-signature and a servicing portal.
• Security and audit: $25K–$70K for SOC 2 Type II, penetration testing and a GLBA or UK GDPR data programme. Sponsor banks will not sign without this.
• Acquisition and test vintages: $40K–$150K to originate enough loans to produce a loss curve worth showing anyone.

Funding routes worth naming in the plan

Digital lending sits awkwardly across funding categories, so be deliberate about which door you knock on:

• SBA 7(a) (US): available up to $5 million, but note that lenders are cautious about financing a business whose core activity is extending credit. It is a more realistic route for the software vendor model than for a lender.
• UK Start Up Loans: up to £25,000 per director at a fixed 6% rate. Useful for a pre-authorisation software or brokerage build, not for a loan book.
• Seed equity: the standard route. Typical asks run $1.5M to $3.5M for 18 to 24 months of runway through first vintages and licensing.
• Warehouse facility: a credit line secured against originated loans, advancing 80% to 90% of principal. Requires a track record or a strong sponsor.
• Forward-flow purchase agreement: a committed buyer for your originations. Often available earlier than a warehouse line and far less dilutive than equity.

Separate your equity ask from your debt ask explicitly. A very common failure in first-draft plans is a single blended "funding required" figure that mixes operating capital with loan capital. Lenders and equity investors are different audiences with different return expectations, and merging the two suggests the founder has not spoken to either.

Three Models, Three Different Companies

"Digital lending platform" describes three businesses that share a category name and almost nothing else. They have different capital requirements, different regulators, different margin structures and different buyers. Pick one before writing, and say which one you picked in the executive summary.

The valuation row is the one founders should read twice. Two companies with identical revenue will be priced very differently depending on which column they sit in, because a software vendor's revenue carries no credit risk and no regulatory tail. That is why so many lending founders eventually try to reposition as infrastructure providers, and why doing so credibly requires the product to have been built for third parties from the start rather than retrofitted.

A hybrid is possible and increasingly common: originate on a bank partnership, prove the credit model on real vintages, then license the decisioning engine to other lenders. If that is your plan, sequence it explicitly with dated milestones. An investor will not assume it.

Take Rates & Unit Economics

Revenue for a fee-based platform is straightforward to describe and easy to get wrong in a model. Every time a loan funds through the platform, the platform earns a fee, usually a percentage of the funded amount, and a smaller ongoing servicing or trail fee for as long as the loan performs (Lendflow). Typical origination take rates sit between 2% and 6%, with consumer unsecured at the higher end and secured or large-ticket commercial at the lower end.

Revenue lines to model separately

• Origination fee: percentage of funded principal, recognised at funding. Your primary line.
• Servicing fee: typically 0.5% to 1.25% annually on outstanding balances. Small per loan, but it is recurring revenue and it materially improves valuation.
• Referral or fallback revenue: fees from routing declined applicants to secondary lenders. Frequently ignored, and often 8% to 15% of total revenue, because most applicants are declined.
• Ancillary products: payment protection, credit monitoring, card products. Regulatory scrutiny here is high, so model conservatively and disclose clearly.
• Software licensing: if you sell decisioning or origination technology to other lenders, model this as a separate SaaS line with its own gross margin, not blended into lending revenue.

A worked example: 1,000 funded loans a month

Numbers below are an Avvale planning model for a US consumer platform on a bank partnership, sized at 1,000 funded loans per month. They are illustrative, but the structure is the one your financial model should mirror.

At 1,000 funded loans a month that is $83,720 of monthly contribution. Set against fixed costs of roughly $155,000 a month for compliance, legal, licence maintenance and engineering, the business is losing about $71,000 a month. Break-even arrives at approximately 1,850 funded loans per month, which is around $27 million of monthly origination volume.

That number is the most useful single output of the whole exercise, and almost no first-draft plan contains it. It tells an investor exactly what the raise has to buy: not "growth", but a specific volume threshold, and how many months of runway are needed to reach it.

The line most plans get wrong

Look again at the credit data line. It costs $111.78 per funded loan even though the per-application cost is only $9.50, because roughly 11.8 applications are screened for every loan that funds. The same distortion applies to marketing. Founders routinely quote a $22 cost per application and present it as customer acquisition cost. At an 8.5% funnel that is really $259 per funded loan, and if a reviewer spots this substitution, everything downstream in the model becomes suspect.

Model the funnel explicitly: applications, approval rate, acceptance rate, funded loans. Then derive every per-unit cost from the funded number. If your product is small-ticket, note that the fixed per-application data cost may make low loan sizes structurally unprofitable regardless of how good the credit model is. That is a real constraint and stating it demonstrates command of the business.

Credit loss, if you hold risk

If you carry any credit exposure, whether a full balance-sheet book or a first-loss piece under a forward-flow deal, the plan needs a loss curve, not a loss percentage. Show cumulative net loss by months-on-book for each origination vintage. A single "5% expected loss" assumption tells a credit investor that nobody on the team has managed a portfolio. Vintage curves, even modelled ones with stated assumptions, tell them the opposite.

Licensing: US, UK, India

Regulation is the moat and the bottleneck in this business. It is also the section most first-draft plans handle in a single vague paragraph, which is a fast way to lose a sophisticated reader. Below is what needs to be specific.

United States

Most non-bank lenders offering credit to consumers must be licensed in each state where they operate. State banking and financial services departments set the obligations, which commonly include background checks, minimum net worth standards, a compliance programme and periodic reporting (CSBS).

• NMLS filing. Most states license through the Nationwide Multistate Licensing System. You create a company record, complete the MU1 form and upload supporting documents. States commonly require audited or pro-forma financials, a detailed business plan, and a description of your lending products including rates, terms and your compliance approach. Your business plan is a licensing artefact, not just a fundraising one.
• Surety bonds. Nearly every state requires one, sized to loan volume and business type, commonly $25,000 to $100,000 per state. Budget the premium annually, not once.
• Cost and timeline. Application fees typically run $500 to $3,500 per state. Individual approvals take 60 to 180 days. A serious multi-state programme is a 12 to 24 month project, and reaching all states is rarely worth it. Most platforms cover 60% to 70% of the addressable population with 12 to 18 states.
• New York. Treat separately. The Department of Financial Services operates its own licensed lender regime with distinct thresholds and a demanding review (NYDFS).
• Money transmitter licences. Required if you hold or move borrower funds rather than routing them bank-to-bank. Bonding and net worth requirements are materially higher than a lender licence. Structure your money flows to avoid this if you can.
• Federal supervision. The CFPB supervises non-bank lenders offering installment loans, BNPL and lines of credit, enforcing TILA, ECOA and FCRA. Examiners focus on disclosures, fair lending, complaint management and data accuracy (InnReg).

United Kingdom

The UK replaces 50 licensing processes with one, but the standard of that single process is high.

• Consumer credit authorisation. Apply to the FCA for full permission if you are lending. Limited permission covers a narrower set of activities. Covered activities include consumer lending, credit broking, debt collection, debt adjusting, debt counselling and operating an electronic system in relation to lending, which is the peer-to-peer platform permission (FCA).
• Timeline. The FCA must decide within six months of a complete application, or twelve months if the application is incomplete. The six-month clock starts when the application is judged complete, not when you submit. In practice, budget 3 to 6 months for limited permission and 6 to 12 months for full permission.
• Fees. The application fee is non-refundable and scales with your pricing category, which depends on the regulated activities you request. Annual periodic fees follow authorisation.
• Prudential standards. P2P platforms carry capital requirements set out in the FCA's crowdfunding policy statement.
• Conduct. The CONC sourcebook governs creditworthiness and affordability assessment, arrears handling and forbearance. Consumer Duty requires evidenced good outcomes across the product lifecycle, and the FCA expects to see the framework at authorisation (FCA guidance for consumer credit firms).

India

India runs the most prescriptive digital lending regime of the three, and it is worth reading even if you never operate there because other regulators are borrowing from it. The Reserve Bank of India (Digital Lending) Directions, 2025 were introduced on 8 May 2025 (Lawrbit, 2025). Key provisions:

• App registration. Regulated Entities must report every Digital Lending App, their own or their Lending Service Provider's, on the RBI CIMS portal, with an initial deadline of 15 June 2025.
• Multi-lender transparency. From 1 November 2025, an LSP working with multiple lenders must display a digital comparison of all matched loan offers, list the lenders that did not match, and avoid dark patterns that bias borrower choice.
• Fee direction. LSP fees are paid by the Regulated Entity. They cannot be charged to the borrower.
• Data minimisation and localisation. LSPs may store only name, address and contact details. All data must be stored in India; anything processed overseas must be deleted and repatriated within 24 hours.
• Non-delegable liability. Outsourcing does not shift obligation. The Regulated Entity remains fully responsible for the acts and omissions of its LSPs and apps, however much is delegated. Both the Regulated Entity and any interfacing LSP must appoint nodal grievance redressal officers.

The through-line across all three jurisdictions is the same: regulators are steadily pushing accountability back onto whoever holds the licence, no matter who built the app or bought the traffic. If your model depends on being the party that touches the borrower but not the party that carries the obligation, that gap is closing. Say in your plan how you have priced the compliance function accordingly.

Six Ways These Plans Fail

Patterns we see repeatedly in digital lending plans that come to us for a rewrite after a rejection.

1. The licensing programme is missing from the budget and the timeline

The plan costs the engineering build precisely and disposes of licensing in a sentence. In reality a multi-state programme is often the largest pre-revenue expense and always the longest lead time. Show a dated state-by-state roadmap with per-state cost and expected approval month, and tie your revenue ramp to it. A reviewer who has financed lenders before will look for exactly this and will notice its absence immediately.

2. Origination revenue is forecast without a credit loss curve

Revenue rises smoothly and losses appear as a flat percentage of the book, or not at all. Losses in a new lending business are front-loaded by vintage and vary sharply by acquisition channel. Present cumulative net loss by months-on-book, split by vintage, with stated assumptions. If you have no data yet, model it from a published comparable and say which one you used.

3. The bank partnership is assumed rather than evidenced

Plans routinely say "we will partner with a sponsor bank" as though it were a procurement decision. Sponsor banks run months-long diligence covering your compliance programme, security posture, model documentation and capitalisation, and they turn founders down regularly. Name the bank if you have one, include the term sheet, and if you do not, show the shortlist, the diligence requirements you are working against, and your target signing date.

4. State rate caps and true-lender risk are not addressed

A plan showing a 29% APR product marketed nationally, with no discussion of state usury caps or which entity bears economic risk on the loan, signals that nobody has taken legal advice. Set out your state-by-state rate matrix, the retained interest of the partner bank, and your legal position. This section may feel defensive. To an institutional investor it reads as competence.

5. The underwriting model has no fair-lending documentation

Machine-learning underwriting is a selling point right up until someone asks how you test for disparate impact and generate ECOA-compliant adverse action reasons for a model that is not directly interpretable. Both your bank partner and the CFPB will ask. Describe your testing regime, your reason-code methodology and who reviews the model.

6. Acquisition cost is quoted per application

Covered above in the unit economics section, and worth repeating because it is the most frequent single error. Cost per application multiplied by the funnel conversion rate gives cost per funded loan, which is usually ten to fifteen times higher. Every plan that gets this wrong overstates contribution margin dramatically, and the correction usually turns a profitable-looking model into a loss-making one.

A seventh, less common but fatal: forecasting a market share percentage of the $303 billion origination figure instead of building volume bottom-up from channel capacity. "We will capture 0.1% of the market" is not a forecast. Applications per month by channel, times approval rate, times acceptance rate, times average loan size, is.

Sample Plan Preview

An extract from a digital lending platform plan built on this template, showing how the sections above resolve into a document a bank partner and an equity investor can both read.

What's in the Template

Every Avvale business plan template includes these sections, pre-structured for your industry:

• Executive Summary — Your business at a glance, written to hook investors in 60 seconds
• Company Overview — Legal structure, ownership, location, and founding story
• Industry Analysis — Market size, growth trends, and the regulatory picture
• Customer Analysis — Target demographics, pain points, and spending patterns
• Competitor Analysis — Competitive mapping and your differentiation strategy
• Marketing Plan — Channels, messaging, and customer acquisition strategy
• Operations Plan — Day-to-day workflows, staffing structure, and key milestones
• Management Team — Founder bios, advisory board, and key hires planned

For a lending business specifically, we recommend adding four appendices the base template leaves optional: a state licensing roadmap with dated milestones and per-state cost, a credit policy summary, a vintage loss curve with stated assumptions, and the funding stack split cleanly between equity and loan capital. Those four are what separate a plan that raises from a plan that circulates.

The optional Financial Forecast add-on (included in our $300/£250 and $1,000/£800 packages) provides a 5-year Excel model with income statement, cash flow, balance sheet, break-even analysis, and startup capital requirements.

Related templates worth reading alongside this one: the micro lending business plan template for smaller-ticket and community lending models, the fintech startup business plan template for the broader category, and the digital banking platforms business plan template if a charter is part of your longer-term plan. Our case study library includes several fintech and lending plans taken to funding.

Frequently Asked Questions

Useful Links & Resources

These links were preserved from the live page so important references and partner links are not lost during the page refresh.

• Facebook
• Instagram
• LinkedIn
• YouTube