CAPITAL MARKETS DATA ARCHITECTURE
Capital markets data,
seen in 20/20.
Where financial data blurs as it moves from trade to settlement, custody, NAV and the regulator, and how to bring it into focus so AI can deliver.
Eighteen years inside that chain, at top-tier global custody and asset servicing institutions. This is a personal collection of case studies, demos and ideas — shared practitioner to practitioner, to read, critique or build on, not as a pitch.
- Focus
- The hand-offs where data blurs: trade to settlement, IBOR to ABOR, custody to NAV, ledger to regulator — across asset servicing, private markets, fund accounting, corporate trust and CLO administration
- 20/20 means
- Every number traced to its source, explained in plain language, and agreed across books
- Currently
- Full-time in capital markets data innovation and solution architecture
- Open to
- Peer conversations, panels, and technical exchange with fellow practitioners
Why 20/20, why now
The struggle is scaling and proving AI on data that does not line up.
Recent surveys of banks, asset managers, private markets firms and post-trade operations point the same way. Firms are not short of AI ambition. They are short of data they can trace, explain and reconcile.
99% / 3%
of financial institutions are piloting AI; 3% have deployed it at enterprise scale. 46% of executives rank better data quality as their top priority.
Ataccama Data Trust Report 2025 · 300 data leaders
64%
of asset managers say AI adds meaningful value only after data quality and structure improve; 65% say fragmented fund data holds back efficiency.
FE fundinfo 2026 · 200 managers
60%
of AI projects without AI-ready data will be abandoned through 2026, Gartner predicts; 63% of organizations lack, or are unsure of, the right data practices for AI.
31%
report advanced data strategy capabilities. “AI is becoming a stress test for enterprise data,” says the EDM Association.
EDMA Benchmark 2026 · 435+ orgs
<10%
of private equity and credit firms report high data maturity; 56% name heavy Excel reliance as a top challenge; about half cite poor system integration.
55%
of industry respondents find it hard to evidence enterprise value from AI; 81% use AI, 14% call it transformational.
51%
of post-T+1 settlement exceptions are still resolved by email or phone; top causes of unmatched instructions are wrong accounts and trade data mismatches.
End to end
lineage from report back to source is now banks' dominant BCBS 239 target, and many lack quality controls on end-user computing.
Deloitte BCBS 239 survey 2026 · 50+ banks
What it means. The pain is not a lack of vision in the strategy sense. It is not being able to see the data clearly as it crosses systems, books and firms — and that is where an optician earns his keep. To be fair to the evidence: BCG's September 2026 index finds nearly half of companies now generate meaningful value from AI (BCG, Sep 2026). The problem is not that AI never pays. It is that firms struggle to scale it, and to prove it.
Figures as reported by each source; each link opens the original publication or a published summary of it. Checked October 2026.
The 20/20 test
Data rarely breaks inside one system. It blurs at the hand-offs between them.
Trade to settlement, IBOR to ABOR, custody to NAV, ledger to regulator — with reference data feeding every step. Eighteen years inside that chain showed me where it blurs, and capital markets is an unforgiving place to learn it: dense regulation, overlapping legacy systems and zero tolerance for a misreported position, a missed NAV tolerance or a broken book of record. Every case study and demo on this site is measured against the same three checks.
Traced
Every number leads back to its source, through every hand-off — the end-to-end lineage regulators now expect banks to show.
Explained
Anyone can say why a number is what it is, and why it moved, in plain language — to a NAV reviewer, an auditor or an AI assistant.
Agreed
The same number in every book: IBOR and ABOR, fund and custodian, ledger and report. Differences are explained, not just flagged.
Get those three right and AI has something to stand on: answers that can be checked, and value that can be proven.
Beyond capital markets
Different industry, same blur.
Capital markets is where these ideas got stress-tested, not where they stop applying. Wherever data crosses systems, vendors or departments, the same test holds: can each number be traced, explained and agreed?
Healthcare & life sciences
Patient, provider and claims data blur at the hand-off between EHR, billing and payer systems — the same shared-vocabulary problem as fund, investor and custody data, with HIPAA in place of the SEC.
Mostly: explained, agreedSupply chain & logistics
Vendor, SKU and shipment records need the same lineage and reconciliation discipline built for security reference data and market data vendors — a bad master record travels just as far.
Mostly: tracedInsurance & risk
Policy, claims and reinsurance data has to agree across legacy platforms — the same books-of-record problem as IBOR and ABOR, with the same reporting and regulatory frameworks built for custody and corporate trust.
Mostly: agreedGovernment & public sector
Multi-agency reporting has to be traced from report back to source — the lineage regulators already ask of banks, and the reporting Centers of Excellence built for capital markets. Different regulator, same architecture.
Mostly: tracedInside the chain
A narrow domain, held deeply.
A lot of data writing borrows capital markets vocabulary from the outside. This starts from the other direction — eighteen years inside the operations of the largest custody and asset servicing platforms, finding where each hand-off blurs and building data structures that keep it in focus in production.
Private markets & alternative investment data
Led Geneva, Investran, and Yardi integrations into ODS/Snowflake for hedge fund and private market clients, and the reporting workstream for credit manager outsourcing implementations.
Blurs at: admin platforms → reporting · Fixes: tracedCustody & corporate trust architecture
Co-designed the data strategy for a $15 trillion custody business and built the IBOR taxonomy spanning trade processing, asset servicing, derivatives, and collateral management.
Blurs at: settlement → custody · Fixes: agreed, tracedCLO & loan servicing automation
Automated security reference data set-up for bank and non-bank debt — cutting processing time from 45 minutes to under 1 minute while reducing operational risk.
Blurs at: vendor feeds → security master · Fixes: tracedRegulatory reporting & data governance
Built reporting Centers of Excellence and data quality frameworks that recovered over $1.8M in annual revenue leakage and $800K in market data costs.
Blurs at: ledger → regulator · Fixes: tracedNAV oversight & intelligence
Design rules-based NAV oversight engines — tolerance checks, exception workbenches, and materiality-scored break management — so no NAV releases without an independent verdict on why it moved.
Blurs at: custody & pricing → NAV · Fixes: explained, agreedInvestment compliance rule architecture
Build tiered compliance rule libraries across regulatory, prospectus, internal, and side-letter sources — with IBOR/ABOR book-of-record integrity as a first-class control, not an afterthought.
Blurs at: IBOR → ABOR · Fixes: agreedWays to engage
How I like to collaborate.
I'm a full-time capital markets professional, not a consultant for hire — this is about trading ideas with other people who work in this space, not scoping paid work.
Peer review & feedback
Happy to read a data model, rule library, lineage design or governance framework you're building, and tell you honestly where I think it blurs.
Speaking & panels
Open to conference talks, webinars, or panel discussions on bringing capital markets data into focus for AI, NAV oversight, IBOR/ABOR integrity, or compliance rule modeling.
Technical writing & case studies
The case studies on this site are public and generalized from real work — open to co-authoring, contributing to industry publications, or just discussing them further.
Mentoring & career conversations
Glad to talk with people earlier in a fund servicing or capital markets data career about how the field actually works day to day.
Field notes
In dialogue with the data community.
None of this happens in a vacuum — it's shaped by, and hopefully contributes back to, the wider conversation happening across the modern data community.
A recurring theme in that conversation, including the Modern Data 101 community, is the shift from treating data as a byproduct of some operational system toward treating it as a Data Product — something with a named owner, a contract describing what it guarantees, a lifecycle, and its own semantics, lineage, and quality signals, discoverable and addressable on its own terms. Market infrastructure firms are already treating it that way commercially — at London Stock Exchange Group, for instance, data quality and data products are core commercial assets, not back-office overhead.
That framing lines up closely with how the NAV oversight and compliance rule work on this site was built — NAV as a versioned, lineage-tracked, entitled deliverable rather than a spreadsheet output; a compliance rule library with explicit data provider attribution and an auditable interpretation log rather than tribal knowledge. Modern Data 101 frames the underlying idea well: a Data Product is "context packaged as a first-class asset" — which is a good description of what a properly governed NAV or reference-data pipeline should be, capital markets or otherwise. In 20/20 terms, a data product is data that arrives already traced, explained and agreed.
Each case study below links to the specific piece of that broader conversation it resonates most with — not as a claim of originality, but as a pointer to where these ideas are being actively worked out by a wider group of practitioners.
Selected work
Projects & artifacts.
Representative outcomes from 18 years across top-tier global capital markets and asset servicing institutions. Each one names the hand-off where the data blurred and the part of the 20/20 test the work fixed. Identifying details are generalized; the structures, systems, and results are real.
Bank & non-bank debt reference data automation
A multi-year project automating set-up and maintenance of security reference data from market vendors — cutting processing time from 45 minutes to under 1 minute and reducing operational risk.
Blurs at: vendor feeds → security master · Fixes: traced
Read the full case study →$15T custody data strategy
Co-designed the data strategy for a global securities and fund services custody division — spanning data governance, quality management, and reference data acquisition.
Blurs at: settlement → custody · Fixes: agreed, traced
Read the full case study →Reporting COE & $1.85M leakage remediation
Built a Center of Excellence for treasury and trade services reporting, normalizing transactional data across systems and jurisdictions — recovering $1.85M in annual revenue leakage.
Blurs at: ledger → regulator · Fixes: traced
Read the full case study →$800K market data cost optimization
Built the infrastructure to track, manage, and allocate market data costs across business units and clients — an $800K annual savings at a global bank's investor services division.
Blurs at: vendor invoices → usage · Fixes: traced
Read the full case study →NAV oversight & intelligence framework
Evaluated rules-based oversight, independent revaluation, shadow NAV reconciliation, and multi-source harmonization approaches, then built a phased NAV oversight rules catalog and rollout roadmap so no NAV releases without an independent verdict on why it moved.
Blurs at: custody & pricing → NAV · Fixes: explained, agreed
Read the full case study →IBOR/ABOR investment compliance rule library
Built a 135+ rule compliance library spanning regulatory, prospectus, internal, and side-letter sources — correcting a common governance blind spot by separating the pre-trade (IBOR) and settled (ABOR) books of record instead of collapsing them into one.
Blurs at: IBOR → ABOR · Fixes: agreed
Read the full case study →Fund servicing & custody ontology — interactive view
A simplified, illustrative render of a canonical data model spanning fund administration, custody, and reference data — the kind of structure behind the security reference data automation and IBOR taxonomy work above. Drag any node to explore; hover for detail.
Illustrative model shown for demonstration — the underlying structure is generalized from real work.
How I work
The approach behind the case studies.
Enterprise data work fails when it stays abstract. Every case study on this site follows the same discipline, and ends with the same check: can each number be traced, explained and agreed?
Find the hand-off where it blurs
Not a generic model — the specific hand-off, decision, or system where the data stopped lining up.
Work from operations, not slideware
Time with the people who touch the data daily — operations, compliance, servicing — before any model gets drawn.
Design for more than one audience
A model that satisfies engineering but loses the business, or vice versa, doesn't survive contact with the org chart.
Leave it in focus
Documentation, lineage, and a coverage matrix a team can maintain, so the data stays traced, explained and agreed after the project ends — not a deck that gets shelved.
Open to conversation
Let's compare notes.
I'm a full-time capital markets data professional, and this site isn't a service offering — it's a place to share work and hear how other practitioners are approaching the same problems. If something here resonates, or you see the data blurring somewhere I haven't covered, I'd genuinely like to hear about it.
info@dataoptician.com