Wealth Data Sets
Realistic households. None of them real.
To build, demo, and test, you need households that behave like a real book of business. Production data drags a privacy review and a DUA behind it, and Faker falls apart the moment your engine sees a 28-year-old with $4M and no income. Wealth Data Sets is neither: 1,451 synthetic households across 71 archetypes, calibrated to behave like the real thing, with zero PII and no data-use agreement. 61 purpose-built Data Sets, or the full Master Corpus, yours to keep.
{
"household_id": "syn_A-01_0017",
"archetype": "young-family-first-home",
"life_stage": "Accumulation",
"demographics": {
"household_type": "married_joint",
"primary_age": 42,
"dependents": 2,
"state": "CA"
},
"net_worth": {
"total": 312500,
"liquid": 38000,
"retirement": 165000
},
"pii": null
}Every Data Set ships in both by default; Parquet on request.
Field derivations, statistical calibration, and integration guidance.
Synthetic by construction, with no privacy review and no data-use agreement.
One-time purchase, royalty-free for internal build & test.
Households to test on, decisions to grade against, evals to prove it.
Six families of household Data Sets, each calibrated for a specific workflow — plus the decision-memory packs that grade agents on them, and the eval packs that prove your AI runs on this year's rules.
Compliance & Audit
Audit-ready households for Reg BI, fiduciary review, fee benchmarking, and lending fairness. Built for compliance teams, supervisory testers, and RIA onboarding.
Reg BI Suitability Audit Pack
130 synthetic households tuned for Reg BI suitability testing — concentrated holdings, age 75+, recent inheritance, cognitive decline markers, and risk-mismatch flags. Each record carries the eligibility triggers required to exercise broker-dealer supervisory workflows end to end.
Cash Flow Stress Test Dataset
270 households with monthly cash-flow models including income shocks, expense spikes, and liquidity-stress scenarios. Variable-income earners, single parents, gig workers, distressed mortgages, and post-divorce rebuild cases. Each household has 96 monthly snapshots.
Fiduciary Fee Benchmark Dataset
130 affluent and HNW households with detailed fee structures: AUM-based advisory fees, tiered breakpoints, fund expense ratios, transaction costs, and tax-drag estimates. Includes complex fee arrangements (multi-firm, family-office, performance-based).
RIA Onboarding Stress Test Pack
400 prospect households covering RIA client variety from formation through retirement. KYC-complete records, goal-based planning fields, initial recommendation outputs, and CRM-compatible field naming. The broadest single bundle by archetype coverage.
CRA / Underserved Lending Pack
100 households underserved by traditional financial services: thin-credit-file ITIN filers, post-bankruptcy rebuilders, low-income working families, recent immigrants, disability claimants, and cannabis-industry workers. Useful for CRA-aligned lending and inclusion analytics.
CDFI / Underbanked Pack
80 underbanked and underserved households: prepaid-card users, check-cashing customers, ITIN filers, post-bankruptcy unbanked, and cash-economy participants. Companion to B23 (CRA) but focused on transactional / banking inclusion rather than lending.
Tax Planning
Tax-complex households for harvesting algorithms, equity comp, K-1 pass-throughs, QSBS, crypto, and multi-state planning. Field-level cost basis, lots, and tax events.
Tax-Loss Harvesting Simulator
350 households with detailed taxable brokerage positions, lot-level cost basis, unrealized gain/loss schedules, and wash-sale tracking. Specific-ID lots, holding periods, QSBS attestations, and cross-account wash-sale flags — built for direct-indexing engines and TLH backtesters.
Equity Compensation Tax Pack
150 households with detailed equity compensation: RSU vesting calendars, ISO/NQSO grants, ESPP with lookback, 83(b) elections, AMT exposure, and exercise window expirations. Each grant has a structured grant_type, vesting schedule, and vested-to-date calculation.
Small Business / K-1 Tax Pack
140 small business owner households across LLC, S-Corp, and partnership structures. Each carries reasonable-salary calculations, K-1 distributions, QBI deduction modeling (with SSTB classification and phaseout), guaranteed payments, and capital account tracking.
Founder & QSBS Tax Planning Pack
60 founder and early-employee households with QSBS Section 1202 attestation chains, 5-year holding period satisfaction, gross-asset tests at issuance, and stacking via non-grantor trusts. Includes IPO and secondary-sale liquidity events.
Crypto & DeFi Tax Pack
50 crypto-heavy households with detailed wallet inventories, transaction histories, cost-basis lots (FIFO/specific-ID), DeFi positions (LP, staking, lending), NFT income, airdrops, and 1099-DA reconciliation. Self-custody and exchange-custody both represented.
Multi-State Tax Optimization Pack
320 households with multi-state tax exposure: HCOL-to-LCOL relocations, dual residency, MA millionaires tax, NY/CA convenience-of-employer rules, and digital nomads with no fixed domicile. Each carries source-state allocation history and residency change events.
Retirement & Healthcare
Near-retiree, retiree, and benefits households with multi-source income modeling: Social Security, pensions, RMDs, IRMAA, HSAs, and withdrawal sequencing.
Retirement Income Sequencing Pack
210 near-retiree and retiree households with multi-source income modeling: Social Security claiming ages, pension elections, RMD schedules, annuity payouts, and taxable/tax-deferred withdrawal sequencing. Includes Roth conversion windows and IRMAA tier calculations.
Healthcare Benefits & HSA Pack
250 households across the healthcare-benefits lifecycle: HDHP-with-HSA accumulators, COBRA/ACA-marketplace gap-fillers, SSDI/LTD claimants, Medicare-bridge pre-retirees, and IRMAA-exposed retirees. HSA-as-retirement-account strategies fully modeled.
Estate & Wealth Transfer
Estate, trust, and wealth-transfer households for legacy planning, dynasty trusts, beneficiary stress tests, and life-event transitions.
Estate Planning Beta Pack
200 households across the wealth-transfer lifecycle: young families establishing wills, dual-income couples buying first homes, peak earners building trusts, retirees coordinating beneficiaries, and blended families with QTIP planning. Each carries a current estate document inventory.
Wealth Transfer Readiness Assessment
110 HNW and UHNW households with estate planning readiness scores, trust structures, gifting histories, charitable giving data, and GST exemption tracking. Complements B09 (Next-Gen Attrition) and B12 (Estate & Trust Planning).
Estate & Trust Planning Pack
180 households representing the full estate-planning lifecycle: peak earners building first trusts, HNW with complex multi-trust structures, late-retirees simplifying for handoff, recent inheritors navigating step-up basis, and next-gen heirs receiving distributions.
Life Transitions Pack
220 households mid-transition: divorce in progress, post-bankruptcy recovery, medical-debt crisis, sandwich-generation caregivers, recent windowhood, sudden wealth, distressed mortgages, and blended-family formation. High behavioral-event density and asset reshuffling.
Risk & Underwriting
Underwriting, illustration, and stress-test households with insurance schedules, mortgage exposure, claims history, and cybersecurity risk overlays.
InsurTech Illustration Validator
240 households with detailed insurance needs analysis: life, disability, LTC, and annuity illustration inputs. High deviation surfaces edge cases (sub-standard underwriting, dependent special needs, blended families, recent claims history) that stress illustration engines.
Mortgage Stress Test Pack
90 households spanning the mortgage lifecycle: first-time buyers stretched to DTI limit, young families with new mortgages, distressed homeowners evaluating modification, and underwater scenarios. Includes complete mortgage application data, current LTV, and reserve adequacy.
Insurance Claims & Cybersecurity Risk Pack
50 households with embedded fraud, claims, and cybersecurity scenario flags: medical claims disputes, account takeover patterns, predatory lending exposure, and disability claim documentation. High deviation for adversarial ML training.
Client Segments
Specialized client segments with distinct planning needs: ESG, next-gen, student debt, military, faith-based, gig income, and behavioral coaching cohorts.
Alt Investment Suitability Dataset
230 accredited and qualified-purchaser households with private equity, hedge fund, real assets, and private credit exposure. SEC accreditation status, illiquidity tolerance scoring, capital-call schedules, vintage-by-vintage TVPI/DPI/IRR for PE positions.
ESG Values Alignment Auditor
25 households tagged for the conditional ESG / values overlay (one of two bundles where race/ethnicity and religion fields are populated, per the v4 privacy contract). MSCI ESG taxonomy alignment, sector exclusion lists, impact allocation percentages, and DAF holdings.
Next-Gen Attrition Predictor Dataset
15 next-generation wealth-recipient households (teen/young-adult heirs from HNW families) with engagement metrics, custodial account structures, kiddie-tax modeling, and intergenerational communication scoring. Companion pack to B11 (Wealth Transfer Readiness).
Student Debt Optimization Dataset
180 households with detailed student loan data: loan types, servicers, IDR plan enrollment, PSLF qualifying-payment counts, refinancing history, and forgiveness-tax-bomb projections. Includes Parent PLUS borrowers and double-consolidation paths.
Impact & Faith Investment Pack
50 households with values-driven investment overlays: ESG screening, halal/biblically-responsible exclusions, impact bond allocations, DAF management, and shareholder advocacy participation. Companion to B08 (ESG Values Alignment) — fully populates the demographic conditional overlay.
Military & Veterans Financial Pack
90 active-duty, retired, and veteran households spanning enlisted to officer ranks. BRS pension elections, TSP allocations, VA loan entitlements, BAH/BAS allowances, SBP elections, VA disability compensation, CRDP/CRSC, and PSLF for nonprofit-employed veterans.
Faith-Based & International Households Pack
70 households with international or faith-based financial considerations: H-1B visa-dependent earners, recent immigrants navigating ITIN-to-SSN transitions, FBAR/FATCA filers, and faith-based investors with strict exclusions. Conditional populates religion field per PLAN privacy contract.
Gig & Variable Income Pack
100 households with non-W-2 income patterns: gig workers, creators, freelancers, royalty-income artists, military with variable allowances, and remote-worker digital nomads. Variable cash flow modeling, quarterly estimated tax payments, self-employed retirement plans.
Behavioral Finance Coaching Pack
130 households with elevated behavioral risk: financial-anxiety avoiders, overconfident DIY investors with concentration, lifestyle-inflation spenders, post-divorce rebuilders, and student-loan tax-bomb candidates. Each carries a behavioral profile and intervention-trigger flags.
Decision Memory
Deterministic advisor-decision episodes over synthetic households: trigger, options, recommendation, override, outcome, and cited rule keys — known-answer evaluation data for AI agent memory.
Reg BI Suitability Audit Pack: Decision Memory Pack
137 advisor-decision episodes over 277 synthetic households — the same 6 client archetypes as the Reg BI Suitability Audit Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Tax-Loss Harvesting Simulator: Decision Memory Pack
282 advisor-decision episodes over 230 synthetic households — the same 10 client archetypes as the Tax-Loss Harvesting Simulator. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Retirement Income Sequencing Pack: Decision Memory Pack
254 advisor-decision episodes over 530 synthetic households — the same 10 client archetypes as the Retirement Income Sequencing Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Cash Flow Stress Test Dataset: Decision Memory Pack
289 advisor-decision episodes over 176 synthetic households — the same 10 client archetypes as the Cash Flow Stress Test Dataset. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Alt Investment Suitability Dataset: Decision Memory Pack
282 advisor-decision episodes over 238 synthetic households — the same 10 client archetypes as the Alt Investment Suitability Dataset. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Estate Planning Beta Pack: Decision Memory Pack
282 advisor-decision episodes over 275 synthetic households — the same 10 client archetypes as the Estate Planning Beta Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Fiduciary Fee Benchmark Dataset: Decision Memory Pack
170 advisor-decision episodes over 192 synthetic households — the same 6 client archetypes as the Fiduciary Fee Benchmark Dataset. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
ESG Values Alignment Auditor: Decision Memory Pack
57 advisor-decision episodes over 36 synthetic households — the same 2 client archetypes as the ESG Values Alignment Auditor. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Next-Gen Attrition Predictor Dataset: Decision Memory Pack
28 advisor-decision episodes over 12 synthetic households — the same 1 client archetypes as the Next-Gen Attrition Predictor Dataset. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
InsurTech Illustration Validator: Decision Memory Pack
285 advisor-decision episodes over 168 synthetic households — the same 10 client archetypes as the InsurTech Illustration Validator. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Wealth Transfer Readiness Assessment: Decision Memory Pack
226 advisor-decision episodes over 245 synthetic households — the same 8 client archetypes as the Wealth Transfer Readiness Assessment. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Estate & Trust Planning Pack: Decision Memory Pack
306 advisor-decision episodes over 427 synthetic households — the same 12 client archetypes as the Estate & Trust Planning Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Mortgage Stress Test Pack: Decision Memory Pack
115 advisor-decision episodes over 64 synthetic households — the same 4 client archetypes as the Mortgage Stress Test Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
RIA Onboarding Stress Test Pack: Decision Memory Pack
340 advisor-decision episodes over 279 synthetic households — the same 12 client archetypes as the RIA Onboarding Stress Test Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Insurance Claims & Cybersecurity Risk Pack: Decision Memory Pack
87 advisor-decision episodes over 57 synthetic households — the same 3 client archetypes as the Insurance Claims & Cybersecurity Risk Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Equity Compensation Tax Pack: Decision Memory Pack
283 advisor-decision episodes over 183 synthetic households — the same 10 client archetypes as the Equity Compensation Tax Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Small Business / K-1 Tax Pack: Decision Memory Pack
282 advisor-decision episodes over 202 synthetic households — the same 10 client archetypes as the Small Business / K-1 Tax Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Healthcare Benefits & HSA Pack: Decision Memory Pack
324 advisor-decision episodes over 391 synthetic households — the same 12 client archetypes as the Healthcare Benefits & HSA Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Student Debt Optimization Dataset: Decision Memory Pack
315 advisor-decision episodes over 204 synthetic households — the same 11 client archetypes as the Student Debt Optimization Dataset. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Founder & QSBS Tax Planning Pack: Decision Memory Pack
113 advisor-decision episodes over 104 synthetic households — the same 4 client archetypes as the Founder & QSBS Tax Planning Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Crypto & DeFi Tax Pack: Decision Memory Pack
85 advisor-decision episodes over 49 synthetic households — the same 3 client archetypes as the Crypto & DeFi Tax Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Multi-State Tax Optimization Pack: Decision Memory Pack
338 advisor-decision episodes over 259 synthetic households — the same 12 client archetypes as the Multi-State Tax Optimization Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
CRA / Underserved Lending Pack: Decision Memory Pack
293 advisor-decision episodes over 172 synthetic households — the same 10 client archetypes as the CRA / Underserved Lending Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Impact & Faith Investment Pack: Decision Memory Pack
114 advisor-decision episodes over 98 synthetic households — the same 4 client archetypes as the Impact & Faith Investment Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Military & Veterans Financial Pack: Decision Memory Pack
189 advisor-decision episodes over 259 synthetic households — the same 7 client archetypes as the Military & Veterans Financial Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Faith-Based & International Households Pack: Decision Memory Pack
174 advisor-decision episodes over 97 synthetic households — the same 6 client archetypes as the Faith-Based & International Households Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Life Transitions Pack: Decision Memory Pack
368 advisor-decision episodes over 483 synthetic households — the same 14 client archetypes as the Life Transitions Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Gig & Variable Income Pack: Decision Memory Pack
225 advisor-decision episodes over 105 synthetic households — the same 8 client archetypes as the Gig & Variable Income Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
CDFI / Underbanked Pack: Decision Memory Pack
176 advisor-decision episodes over 101 synthetic households — the same 6 client archetypes as the CDFI / Underbanked Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
Behavioral Finance Coaching Pack: Decision Memory Pack
259 advisor-decision episodes over 176 synthetic households — the same 9 client archetypes as the Behavioral Finance Coaching Pack. Each episode records what triggered the decision, the options weighed, the recommendation, any override and its reason, the outcome, and the IRS/Treasury figures that governed the choice (cited by Rule Sets fact key). Deterministically generated, so every QA task ships with a ground-truth answer key — evaluate an agent-memory pipeline against answers known by construction, not labeled after the fact.
DecisionSynth Full Decision Corpus
The complete commercial decision corpus: 1,979 deterministic advisor-decision episodes with 4,913 QA evaluation tasks and answer keys over 1,752 fresh synthetic households spanning all 71 client archetypes. Every episode records trigger, options, recommendation, override, outcome, and cited IRS/Treasury rule keys — ground truth known by construction, version-stamped in _meta, byte-identical on regeneration. Disjoint by construction from the free DecisionSynth Bench dev set and the private held-out set.
Grade the machine before it advises anyone.
Tax-year-vintaged eval packs with answer keys by construction — required figures cited to the establishing IRS document, wrong-but-plausible stale values enumerated with reason codes, scored mechanically. Not household data: proof.
Rule-Grounding Eval Pack (TY2026)
156 tasks · keys + forbidden tables · scorer
Threshold & Cliff Eval Pack (TY2026)
95 tasks · 36 adversarial flips · scorer
Decision-Recall Eval Pack
150 tasks · episodes included · dual delivery
AI Eval Sets Full Corpus
401 tasks · every family · exclusions documented
Try before buying: the 50-task open split is free — /benchmark · line overview at /ai-eval-sets
Master Corpus
The full corpus — every household, every overlay, every longitudinal snapshot. The single deliverable that powers all other Data Sets.
from $495
61 Wealth Data Sets, or the full Master Corpus. The corpus is static, so you own it. No subscription.
Frequently asked questions
Is there an API to pull the full corpus?+
No. The Wealth Data Sets corpus is a one-time purchase delivered as files after checkout, not a metered data API. The public endpoints (the /api/for-agents manifest and one sample household per archetype) are deliberately insufficient for production: PII removed, dollar amounts banded, the 96-month trajectory truncated to 12 months, and projections withheld. They demonstrate fidelity only. The validated corpus is the paid product.
Does the data contain any real customer data or PII?+
No. Every household is synthetic by construction, so there is nothing to anonymize and nothing to leak. There is zero PII and no data-use agreement required, which is why the data can go into demos, CI, model training, and compliance walkthroughs without a privacy review.
How realistic are the synthetic households?+
Each household is a complete, internally consistent financial life across income, assets, debts, insurance, goals, and a 96-month trajectory, calibrated to public sources such as Census ACS, the Federal Reserve Survey of Consumer Finances, BLS, and SSA. A free watermarked sample lets you run real records through your own code and judge fidelity before you buy.
What formats do I get, and how are they delivered?+
JSON and CSV by default, with Parquet on request, plus a methodology document. Files are delivered after a one-time Stripe checkout and are yours to keep. There is no subscription for the static data.
Can I try before buying?+
Yes. A free watermarked sample is available from the sample page, and one deliberately-insufficient evaluation household per archetype is exposed at /api/for-agents for machine consumption. The full validated corpus, with exact figures, complete trajectories, and projections, is the one-time purchase.
Is this a subscription?+
No. The households and decision episodes are static, so they are a one-time purchase you own. Only the separate Rule Sets planning-figures feed, whose values change with the tax code, is a subscription. Rent what changes; own what does not.