Guide to Income-Share Agreement underwriting data for coding bootcamps.

An Income-Share Agreement (ISA) is a financing instrument under which a student receives education funding in exchange for an obligation to pay a fixed percentage of post-program income for a defined number of months, subject to an income floor below which payments are paused and an aggregate cap that limits total repayment. The instrument was popularized in the bootcamp segment by Lambda School (later Bloom Institute of Technology) starting around 2017 and adopted in various forms by App Academy, BloomTech, Holberton, Make School, Pursuit, and roughly fifty other operators at the segment’s 2019-2021 peak. The model collapsed substantially after 2021 due to a combination of regulatory pressure, default rates higher than initial pricing assumed, and operator-specific failures. The Consumer Financial Protection Bureau (CFPB) issued a 2021 Consent Order against BloomTech (then known as Lambda School) finding that ISA contracts that were not described to students as credit and that did not comply with TILA and other lending laws constituted unfair, deceptive, or abusive acts or practices. The Consent Order required BloomTech to refund affected students and reformulate its disclosure practices, and the CFPB has subsequently treated ISAs as a credit product subject to TILA disclosure, ECOA fair-lending requirements, and state lending licensure where applicable. The 2026 bootcamp ISA market is smaller, more carefully structured, and dominated by operators that either have CFPB-aligned disclosures and bona fide credit-product treatment or have shifted to deferred tuition (a similar payment structure without the equity-like income-share feature) or hybrid finance partners (a CFPB-licensed lender originating the credit with the bootcamp providing servicing support). The underwriting question — who should receive the funding — is shared across these structures even where the legal wrapper differs.

Empirical analysis of bootcamp ISA performance through 2023, drawn from public CFPB filings, court records in BloomTech-related litigation, and industry analyses by the Career Education Colleges and Universities (CECU) and individual operator disclosures, identifies four primary drivers of default rate. First, program completion — students who do not complete the program face significantly worse outcomes regardless of financing structure. Bootcamp completion rates vary widely (sixty-five to ninety percent across operators) and pre-cohort fit signals predict completion. Second, post-program placement — students who complete but do not place into a qualifying job (typically defined as software-engineering, data, or related role above an income threshold like $50,000) trigger the ISA income floor and produce reduced repayment. Placement is the dominant driver of ISA economics for completers. Third, post-placement persistence — students who place but exit the role within twelve months produce reduced repayment relative to the modeled trajectory. Fourth, macroeconomic conditions — the 2022-2024 tech-hiring contraction substantially worsened ISA economics across the segment because new-grad placement rates declined irrespective of program quality. From an underwriting perspective the controllable drivers are the first three. Pre-cohort signals that predict completion (academic background, prior coding exposure, program-fit assessment) and post-program placement (interest-aptitude alignment, work-readiness traits) reduce default risk. Post-placement persistence is largely driven by employer-fit and personal factors that are harder to underwrite, though work-readiness and conflict-style assessments give partial signal. Pricing the ISA correctly given the underwritten risk is then a separate exercise; underwriting and pricing should be distinct components of the model.

Career-assessment platforms fit ISA underwriting in two specific ways and explicitly do not fit in others. They fit as predictors of program completion. RIASEC interest profile, work-readiness traits (conscientiousness, EQ, conflict styles), and self-reported skills baseline correlate with completion in published education research; the same correlation appears in bootcamp-specific data analyses. A student whose interest profile is mismatched to a coding-program destination, whose conscientiousness is below the program’s observed completion threshold, or whose self-rated skills baseline is far from program prerequisites is at higher completion risk. Bootcamps using platform data for completion underwriting typically use these signals as inputs to a fit score that informs admissions, with weights determined by the bootcamp’s observed completion data over previous cohorts. They fit as predictors of post-program placement to a more limited extent. The interest-aptitude alignment between the student’s profile and the destination role provides some predictive lift; work-readiness traits provide some lift on the soft-skill side of placement; macroeconomic and employer-side factors dominate. Career-assessment platforms do not fit ISA underwriting as substitutes for technical-skill testing (coding pre-tests like CodeSubmit, HackerRank, or in-house assessments are the relevant tools for technical baseline), as substitutes for academic-background review (academic readiness for the program’s pace), or as substitutes for credit-related underwriting (income, employment, banking history) where the structure is treated as credit. The platform contributes a fit-and-trajectory signal that supplements rather than replaces the existing underwriting components. Bootcamps that have integrated this signal report one to three percentage point reductions in first-cohort default rate after deployment, but with substantial variance across operator size and program model.

When an ISA is treated as credit — which the CFPB has indicated is the default treatment under federal lending law — the underwriting process is subject to the Equal Credit Opportunity Act (ECOA, 15 USC §1691) and Regulation B (12 CFR Part 1002). ECOA prohibits discrimination in any aspect of a credit transaction on the basis of race, color, religion, national origin, sex, marital status, age (provided the applicant has the capacity to contract), receipt of public assistance, or exercise of rights under the Consumer Credit Protection Act. The relevant practical implications for using assessment data in underwriting are threefold. First, the data must not be used in a way that produces a prohibited basis discriminatory effect, including disparate impact under the established Reg B framework. If a particular trait or interest signal correlates with race, ethnicity, gender, or another protected class, using that signal as an underwriting input could produce disparate impact even with no discriminatory intent. The fair-lending response is a thorough disparate-impact analysis on the model and a business-justification analysis on each input — inputs that produce disparate impact must be replaced with less-impactful alternatives where available. Second, the underwriting decision must be explainable to the applicant under the adverse-action notice requirement of Reg B §1002.9 — a denied applicant is entitled to specific reasons for the denial, which means black-box AI scoring is harder to defend than transparent inputs with documented weights. Third, the model must be tested for ongoing fairness as data accumulates; ECOA-aligned model governance is now a recognized practice. Bootcamps using assessment data in underwriting should integrate the platform output as a transparent input with documented business justification, run periodic disparate-impact analysis, and maintain adverse-action notice infrastructure.

A defensible workflow has six stages. First, applicant intake — the prospective student submits an application with academic background, prior experience, financial information where credit treatment applies, and consents required for the underwriting components. Second, technical-baseline assessment — the applicant completes a coding pre-test designed to indicate readiness for the program’s pace; this is the single highest-weight underwriting input for technical bootcamps and is typically delivered through a platform like CodeSubmit, HackerRank, or an in-house instrument. Third, fit-and-trajectory assessment — the applicant completes a career-orientation battery (RIASEC, Skills Audit, Big Five, EQ); the platform produces a fit score against the program’s destination role profile and a completion-risk signal based on prior-cohort data. Fourth, credit-related review where applicable — income, employment, banking history reviewed under TILA and ECOA-aligned procedures. Fifth, underwriting decision — a model combining technical baseline (typically thirty to fifty percent weight), fit-and-trajectory (typically ten to twenty-five percent), and credit indicators (the remainder) produces an approve / decline / approve-at-modified-terms output. Sixth, adverse-action notice and appeal — declined applicants receive notices specifying reasons and an appeal mechanism. The whole workflow is documented for ECOA-aligned model governance and periodic disparate-impact testing. JobCannon’s contribution to the workflow is the third stage. The platform produces per-applicant assessment exports with documented scoring transparency, supports cohort-level analysis to update bootcamp-specific weights as outcome data accumulates, and exports adverse-action-friendly explanations of the score components when needed.

The bootcamp finance landscape in 2026 includes four main structures. ISAs in their classic form continue to operate at a smaller subset of bootcamps with TILA-aligned disclosures and CFPB-compliant servicing. Deferred tuition is structurally similar to ISA but typically with a fixed total payment amount rather than an income share, which simplifies the legal characterization and may avoid some ISA-specific regulation; underwriting concerns are substantially the same because completion and placement remain the dominant repayment drivers. Lender partnerships have become more common, with a CFPB-licensed lender (often Climb Credit, Ascent Funding, or Meritize) originating the credit and the bootcamp providing servicing support; underwriting in this structure is conducted by the lender under their own model, with the bootcamp typically providing program-completion and placement data. Employer-sponsored bootcamp models have emerged where an employer pays for cohort training in exchange for a placement commitment from the trainee — underwriting in this structure is light because the employer assumes the risk, and assessment data primarily supports the employer’s candidate-fit screening rather than a credit decision. Across all four structures the underlying question — who should be admitted — is informed by similar signals. Career-assessment platforms support all four structures with the same fit-and-trajectory signal; they do not directly support credit underwriting in lender-partnership structures, where the lender uses traditional credit data. JobCannon’s production posture supports the candidate-fit signal across all four structures, exports consented per-applicant data to bootcamp underwriters or partner lenders with appropriate authorization, and avoids credit-data ingestion that would put the platform itself in CFPB scope.