Approval rate is one of the most closely watched metrics in digital lending, and one of the least honestly discussed. Lenders report aggregate approval rates that blend prime, near-prime, and thin-file segments in ways that make comparison nearly meaningless. A 68% approval rate at a lender whose applicant pool is 80% prime borrowers and a 68% approval rate at a lender whose applicant pool is 40% thin-file borrowers are very different numbers.
This article is about the benchmarks that actually matter for evaluating underwriting performance — broken down by applicant segment, lending product type, and what the data suggests about where the industry leaves money on the table.
How to Benchmark Approval Rates Meaningfully
The first problem with approval rate benchmarking is segmentation. Aggregate approval rates without segment-level breakdown conflate two fundamentally different performance questions: (1) how well does your model rank risk within a bureau-scoreable population? And (2) what fraction of your thin-file and credit-invisible applicants are you successfully underwriting?
The second problem is that approval rate alone is an incomplete signal. A lender who approves 80% of applicants but has a 12% 90-day default rate is not outperforming a lender who approves 60% with a 4% default rate. The combined metric that matters is risk-adjusted approval rate — the fraction of applicants approved who perform within acceptable loss thresholds.
With those caveats, here's how the industry segments break down:
BNPL: The Highest-Volume Thin-File Segment
BNPL platforms typically have the highest proportion of thin-file and first-time credit users in any digital lending vertical. First-time buyers using BNPL for a $150–$400 purchase are often in their early 20s, recent immigrants, or consumers who've actively avoided credit products. Bureau signal is often minimal or absent.
Industry-reported BNPL approval rates tend to cluster in the 70–85% range — which sounds high. But those numbers include a substantial volume of repeat customers (who have internal repayment history from prior BNPL transactions). First-application approval rates for thin-file applicants with no prior BNPL history are meaningfully lower. Industry observers and CFPB consumer complaint data both suggest that thin-file first-applicants see approval rates 15–25 percentage points below the published aggregate for major BNPL platforms.
The revenue implication: BNPL economics are built on repeat purchase behavior from approved and activated customers. Declining thin-file first-time applicants doesn't just lose the current transaction — it loses the acquisition of a customer whose lifetime purchase volume and repayment history would have been available for better future decisioning. The cost of a false decline in BNPL is higher than the cost of a single declined transaction.
Personal Loans (Digital-First Lenders): The Near-Prime Segment
Digital-first personal loan platforms — offering $1,000–$15,000 unsecured loans — typically serve a near-prime population with FICO scores in the 580–680 range. This is the segment where cash-flow underwriting has the most directly measurable approval rate impact.
Published approval rates at digital personal loan platforms vary considerably by credit appetite — platforms operating at the near-prime end report approval rates in the 35–55% range on their incoming applicant mix. The CFPB's 2023 Consumer Credit Panel data shows that applicants with FICO scores below 620 applying for personal loans have denial rates exceeding 60% at traditional lenders, with digital lenders performing somewhat better but still declining the majority of this segment.
The population that sits just below most digital lenders' FICO cutoffs — typically 580–620 — represents a meaningful fraction of declined applicants who have demonstrably stable cash flows. Federal Reserve Survey of Consumer Finances data suggests that median household income for consumers in this FICO band is not substantially different from the 620–680 band. The difference is the depth of credit file, not the underlying financial stability.
Lenders who supplement bureau scoring with cash-flow features for this near-prime segment typically see approval rate improvements in the 8–18 percentage point range on the 580–640 FICO sub-segment, with default rates that remain within acceptable ranges — because the cash-flow features are doing additional risk stratification within a population that bureau models can only see in aggregate.
Neobank Credit Lines: The Deposit-Customer Advantage
As discussed in our piece on the neobank credit expansion playbook, neobanks have a structural information advantage over external lenders when underwriting their own deposit customers. Yet most neobanks with credit products are underwriting deposit customers with the same bureau-first models they'd use for a cold-acquisition credit applicant.
Approval rates for neobank credit line products — typically $500–$3,000 revolving credit lines offered to existing deposit customers — cluster around 40–55% at platforms using bureau-first underwriting. For deposit customers with 12+ months of account history showing regular activity, those approval rates are substantially understating the credit quality of the available pool.
The differential between what a bureau model sees and what a cash-flow model can see is widest precisely in the deposit-active thin-file segment — the customer who has been banking with you for 18 months, paying rent via Zelle, with consistent payroll deposits and no negative balance events, but a FICO of 610 because they've never had a credit tradeline. Bureau models see 610. Cash-flow models see 18 months of responsible financial behavior.
The False Decline Rate: The Metric Nobody Talks About
False decline rate — the fraction of declined applicants who would have repaid if approved — is perhaps the most consequential metric for evaluating underwriting performance, and it's almost never disclosed. It can't be measured without prospective data from a randomized credit experiment or a shadow portfolio, which makes it practically difficult to compute.
What we can estimate: for a bureau-only model applied to a thin-file applicant pool, the false decline rate is likely in the 25–45% range based on retrospective analysis of thin-file cohorts who were subsequently approved (by different lenders or on reapplication) and performed well. The CFPB has referenced similar estimates in its reports on credit access for underserved communities.
This is the number that should be making product and risk teams uncomfortable. If 30–40% of the thin-file applicants you're declining would have repaid, you're not just leaving approval rate on the table — you're misclassifying customers at a rate that, in a traditional scoring context, would be considered a model performance problem worth prioritizing.
What "Approval Rate Improvement" Actually Means for Revenue
A concrete example to ground the numbers: consider a digital lender processing 8,000 credit applications per month, with 40% coming from applicants with FICO scores below 640 (thin-file or near-prime). Current approval rate: 52% overall, with a 31% approval rate on the sub-640 segment. Expected revenue per approved loan: $340 (origination fee + net interest income on a typical 12-month personal loan).
The sub-640 segment: 3,200 applications per month, 992 approvals at 31% approval rate. An 8-point approval rate improvement on this segment (31% → 39%) through cash-flow underwriting would produce approximately 256 additional approvals per month, or roughly $87,000 in additional monthly revenue at $340 per funded loan.
Against a per-decision API cost in the $0.05–0.10 range, the ROI on the underwriting improvement is straightforward even before accounting for the compound value of customer relationships that would not otherwise have existed.
We're not saying these exact numbers apply universally — product mix, average loan size, APR, and portfolio loss rates vary significantly across lenders. The arithmetic is illustrative, not a guarantee. What the arithmetic does show is that approval rate improvement on the thin-file segment is worth modeling explicitly as a revenue line item, not just a feel-good metric about credit access.
Where the Industry Is Converging
The direction of travel in digital lending underwriting is clearly toward cash-flow and alternative data supplementation of bureau scores, not replacement of bureau scores. The lenders who've invested in building or integrating cash-flow underwriting are not generally abandoning FICO — they're using it as one signal among several, weighted appropriately by its predictive value for a given applicant segment.
The benchmark question for any digital lender today is not "should we use alternative data?" but "how much signal are we currently leaving on the table in the thin-file segment, and what does it cost us per month?" Running that calculation explicitly — against your own application data, your own approval rates, your own revenue per funded account — is the starting point for a realistic evaluation of where alternative data underwriting fits in your stack.
