Financial Innovation Happens in Distribution, Not in Products | HL Hunt
Financial Innovation Happens in Distribution, Not in Products
Strip the branding from consumer financial products and remarkably little has changed. Someone borrows an amount, uses it, and repays over time with a charge for the use. Someone holds a balance and can move it. Someone pays a premium against a contingency. What changes constantly — and where nearly all the effort and capital go — is how those things reach people: at what moment, through whose interface, embedded in what other transaction. That's not a failure of imagination. Product economics are bounded by things that don't move, and distribution isn't bounded by anything — which has a consequence worth sitting with, because it means most of what gets called innovation cannot expand access, only reallocate it.
In this report
The floor under the product
A lender's price can't fall below the sum of what the loan costs to provide:
| Component | Moves with design? |
|---|---|
| Cost of funds | No — set by markets and the lender's own access |
| Expected losses | Marginally, through better selection |
| Origination cost | Yes — and this is where the movement has been |
| Verification cost | Yes |
| Servicing and collection | Somewhat |
| Compliance and capital | No, and rising |
Two of the six yield meaningfully to effort, and neither is the product. They're operations.
Which explains a pattern our selection analysis established without naming this consequence: small loans are scarce because fixed costs exceed what the loan supports, and no product design changes that. You cannot design your way to a viable $300 loan; you can only make originating one cheaper.
And per our definitional analysis, the products that appear genuinely new usually aren't. Earned wage access, lease-purchase, and merchant advances all share the shape of value now for more later — the same economic object with different attributes, arranged to fall outside a category. That's a legal and distributional innovation, not a product one.
You cannot design your way to a viable $300 loan. You can only make originating one cheaper — and that isn't product work.
Why distribution isn't bounded
The asymmetry that determines where effort goes:
- It moves fast. A new channel is a partnership and an integration — months, not years.
- It's unilateral. No regulator sets a floor on how conveniently you may present an offer.
- Returns are immediate and measurable. Conversion, volume, and acquisition cost all move within a quarter.
- It's the point of maximum influence. Per our search analysis, most consumers accept the first adequate offer rather than comparing — so being the offer in front of someone is worth more than being the best offer available.
- It compounds with the asymmetry in our prediction analysis: a firm that predicts behaviour well can identify the moment of maximum receptivity, and distribution is how that prediction is monetized.
Compare a cost-structure improvement: slow, expensive, hard to attribute, and it takes years of loss experience to demonstrate. A rational firm allocates to distribution not because it creates more value but because it produces returns the firm can see.
Which is a version of the measurement problem from our measurement analysis, operating at the level of corporate investment: the returns to distribution are visible and attributable, the returns to cost reduction are diffuse and delayed, and capital follows the measurable one.
Redistribution versus expansion
The central distinction, and the one that determines whether any of this matters to people who currently can't borrow.
| Redistribution | Expansion | |
|---|---|---|
| Mechanism | Reach demand that already existed | Make an uneconomic transaction viable |
| Requires | A better channel or moment | A lower cost of serving |
| Who gains | The firm, versus a competitor | People previously excluded |
| Effect on price | Little | Direct |
| Visible as | Growth | Growth |
The bottom row is why this is hard to see from outside. Both look like a firm serving more customers, and the distinction between capturing a competitor's borrower and serving someone nobody could serve doesn't appear in any volume figure.
The test that separates them: did the cost of serving the marginal customer fall? If not, whoever is now being served was already being served, or was already servable and simply wasn't reached.
This isn't a claim that redistribution is worthless. Reaching someone at a useful moment with less friction is genuine value, and the search analysis suggests it can improve outcomes where it substitutes a reasonable offer for a worse one that would otherwise have been accepted. But it is a claim about who benefits: convenience accrues to people who already had options, and the excluded population is unaffected by a better interface.
What has actually expanded access
Apply the test and a short list emerges. In each case a fixed cost fell:
- Automated decisioning removed underwriting labour from the per-loan cost, which is what made small-balance lending viable at all.
- Source-connected verification removed document handling and manual review, per our verification guide.
- Standardized credit reporting, which per our trust analysis is portable verification — one institution relying on another's accumulated observations rather than establishing the facts directly.
- Payroll-linked provision, which per our employer analysis eliminates verification, collection, and acquisition simultaneously.
- Alternative data, where it genuinely lets a thin-file applicant be assessed cheaply rather than expensively.
- Electronic payment rails, which lowered servicing and collection costs.
Not one of these is a product. They're all reductions in the cost of finding something out or moving money — the two things that make lending expensive.
And the payroll case is the cleanest demonstration, because it holds the product constant. The identical $600 advance to the identical person costs roughly $199 through an ordinary lender and $22 through payroll. Nothing about the product changed. The access change came entirely from where the transaction sat.
Which yields a diagnostic worth applying to any claimed innovation: name the cost that fell. If nothing did, the innovation is a channel — valuable to the firm, neutral for access.
Why embedding is the dominant form
The specific direction distribution innovation has taken, and it follows from the argument.
Financial products increasingly appear inside non-financial transactions — at a checkout, inside a marketplace, within a software platform a business already uses, through an employer. The reasons are exactly the properties above:
- The moment is better. Credit offered when someone is deciding to buy something meets demand at its peak.
- Acquisition cost approaches zero, since the customer is already there.
- Context supplies information — the platform knows what's being bought and frequently the customer's history with it.
- Comparison is at its weakest. Per our complexity analysis, an offer presented at the point of another decision competes against nothing.
That fourth point deserves emphasis because it's the least discussed. Embedding doesn't only reduce acquisition cost — it removes the offer from the comparison set entirely. A borrower who would have compared three lenders for a standalone loan compares none when the credit appears inside a purchase they'd already decided on.
Which is the same escape our complexity analysis describes, achieved by placement rather than by pricing structure. Complexity makes the price unrankable; embedding makes it unranked, and the second is cheaper to achieve and harder to address by disclosure.
And where the embedded offer does supply genuine information advantages — a platform that knows a merchant's transaction history, per our payment data guide — that's a real cost reduction and does expand access. The two effects coexist in the same product, which is why embedded finance resists a single verdict.
What this means about the chain
A conclusion that connects several of this desk's findings and wasn't visible from any of them alone.
Our chain analysis documented six or more firms in a single consumer transaction. Our barbell analysis explained the supply side: mid-sized firms rent fixed-cost infrastructure they can't fund.
This report supplies the demand side. The chain is long because distribution is where value is captured, and distribution is a separable layer.
The logic: if the product is standardized and its economics are fixed, then the party controlling access to the customer captures the surplus, regardless of who holds the risk or funds the loan. That party has no reason to also be a lender, and the lender has no reason to also own the channel. So they separate — and the chain is what the separation looks like.
Which reframes intermediation as a consequence rather than a cause. Chains didn't lengthen because firms specialized; they lengthened because the value sits in a layer that can be owned independently of the product. That predicts the chain's shape: the entities closest to the customer capture the most, the entities holding the risk capture less, and the accountability diffusion our chain analysis documents is a side effect of a separation that happened for entirely different reasons.
The apparent exceptions
Cases that look like product innovation and mostly aren't, worth working through because they're the counterexamples people reach for.
Instalment offerings at checkout. A short-term instalment loan is an old product. What's new is the placement and the merchant-funded economics — a distribution and pricing-incidence change, not a product one. The genuine novelty is who pays, which is real and is still not the product.
Revolving lines with usage-based features. Feature variation on an existing structure. Per our complexity analysis, added dimensions frequently serve comparison resistance rather than function.
Income-share arrangements. The strongest candidate for genuine product novelty, since the repayment obligation is contingent on an outcome rather than fixed — which changes risk allocation rather than just packaging. Worth conceding as a real exception, and worth noting that the categorization difficulties in our definitional analysis apply to it directly.
Secured cards and credit builders. These do expand access, and the mechanism is instructive: they lower the cost of establishing trust rather than changing the product. A secured structure removes the need for expensive assessment, which is a verification-cost reduction — consistent with this report rather than an exception to it.
The strongest objections
"You've defined product narrowly enough to guarantee the conclusion." The serious objection. If a product is only its cash-flow structure, then of course nothing changes. Our response: the definition is doing work but it's the right work, because the question is whether the set of people who can be served gets larger, and that depends on cost rather than on structure. A definition that counted every feature variation as product innovation would obscure exactly the distinction that matters.
"Distribution improvements do reduce costs." Correct and conceded — a channel with near-zero acquisition cost has genuinely reduced a component of the floor, and our own payroll example is the strongest instance. The argument isn't that channels never lower costs; it's that channels which lower costs expand access and channels which merely reach people better do not, and most distribution effort is the second kind. Naming the cost that fell distinguishes them.
"This is unfalsifiable — any counterexample gets recategorized." A fair worry about the structure of the argument. The discipline we'd accept: the test is prospective and specific. Point to a claimed innovation, identify whether a cost component fell, and check whether the population served widened rather than shifted. If access widens repeatedly with no identifiable cost reduction, the thesis is wrong.
Testable implications
- Pricing on comparable products should show little secular decline despite decades of visible innovation, except where an identifiable cost component fell.
- Growth from new channels should show substantial substitution from existing providers rather than net new borrowers.
- Access expansion should be traceable to specific cost reductions in every case, with no unexplained instances.
- Embedded offers should show materially lower comparison behaviour than standalone equivalents, holding the borrower constant.
- Value capture along a chain should concentrate at the customer-facing layer rather than at the risk-holding layer.
- Firms should allocate more to distribution than to cost reduction even where the second has a higher expected return, because of the measurability asymmetry.
The second is the one that would settle the practical question and is answerable from data lenders hold. If borrowers arriving through a new channel were mostly borrowing already, the channel redistributed; if they weren't, something genuinely expanded and the cost that fell should be identifiable. Nobody appears to report this, and it's the number that would tell you whether a decade of innovation reached anyone new.
The conclusion we'd hold: consumer finance has been enormously inventive about reaching people and almost static about what it reaches them with. That's a rational response to which layer moves — and it means the interventions that would widen access are unglamorous operational ones, in the places nobody is competing.
Frequently asked questions
The economics are bounded by funding cost, losses, and operating cost, none of which yield to design. An interface can change who gets the loan, not what it costs to provide.
Mostly it redistributes — capturing demand that already existed, frequently from a competitor. Expansion requires the cost of serving someone to fall below what the transaction supports.
It's the only layer a firm can move quickly and unilaterally, and its returns are immediate and attributable while cost-structure gains are slow and diffuse.
Automated decisioning, source-connected verification, standardized reporting, and payroll-linked provision — all reductions in the cost of finding something out or moving money. None is a product.
Key takeaways
- Only two of the six components of a loan's cost yield meaningfully to effort, and neither is the product — they're operations.
- Distribution absorbs the effort because it moves fast, is unilateral, and produces returns a firm can measure this quarter.
- Redistribution and expansion both look like growth; only a fall in the cost of serving the marginal customer distinguishes them.
- Every genuine access expansion traces to a fixed cost falling — automated decisioning, connected verification, portable reporting, payroll provision.
- Embedding removes an offer from the comparison set entirely, which is cheaper to achieve than pricing complexity and harder to fix by disclosure.
- Chains are long because value sits in a layer that can be owned separately from the product, which makes intermediation a consequence rather than a cause.
This report presents an analytical framework and the authors' interpretation; it is not investment, legal, or financial advice. The distinction between redistribution and expansion depends on data that is not generally published, and no quantification is offered here; the implications identified as testable should be treated as hypotheses.