Practice

Thirty years of investment operations across the front, middle and back office, for hedge funds, insurers and real money portfolios. The work divides into two halves that are really one thing: the infrastructure a desk runs on, and the governance that has to be provable when somebody asks.

Everything on this page points at something you can open and check.

Investment Operations and Technology

Data, systems, workflow and controls. The environment an investment decision has to survive after it is made.

Investment operations at the legacy ceiling

The monthly reporting deck is always late. Daily reconciliations break. Another manual workaround gets added to get through the day. These can look like separate problems. They are usually symptoms of the same one: data that is inconsistent, difficult to access, or arriving too late to be useful.

Most legacy systems did not start out as bad systems. They were built or purchased for good reasons and, for years, they worked.

What changed was the business around them.

More products. More data. More reporting. More controls. More complexity. Eventually a system that made sense ten years ago becomes a constraint nobody deliberately chose. It happens gradually. A workaround becomes a permanent process, a manual reconciliation becomes part of the daily routine, and an exception queue grows until managing it is someone's full-time job.

One of the clearest signs is onboarding. If adding a new asset class, instrument type or data source requires significant engineering effort every time, the infrastructure has reached its ceiling. The same is true when bad reference data gets processed first and reconciled later, after it has already reached valuations, payments or client reporting.

The cost shows up across the operation.

Economic. Maintenance costs rise as technical debt builds, and new business is limited by what the platform can support.

Operational. Manual intervention becomes embedded in core processes, and answering questions from clients, auditors and regulators takes too long.

Governance. Data lineage is difficult to trace, controls are performed by hand, and errors are found downstream instead of stopped at the source.

Talent. Critical knowledge concentrates in a small number of people, while technical staff spend their time maintaining legacy processes instead of improving them.

Asset-class complexity compounds all four. Private credit has limited data standardization and a servicing process that remains highly manual. OTC derivatives carry lifecycle events and economic terms buried in documents. Structured products require hundreds of instrument attributes that general-purpose platforms were never designed to hold. In each case the business introduces complexity the existing architecture was not built to support. Without the right foundation, that complexity gets absorbed through people, spreadsheets and manual processes.

Scale exposes the same problem faster. A platform built for twenty thousand positions does not simply become a little slower at three hundred thousand. Eventually something breaks. And the constraint is rarely trade execution. The harder problem comes before the trade: modeling the instrument correctly, establishing the reference data, and making sure downstream systems can actually process what the investment team intends to own. Often those terms are sitting inside credit agreements and servicing notices that someone is keying in by hand.

That is the real ceiling on new business.

Not capital. Not investment appetite. Whether the operating environment can represent, process, control and report on the investment.

The time to deal with it is before something else forces the decision: a new reporting obligation, a vendor sunsetting the platform, a failure during a close, a mandate you could not support. Once the decision is made against a deadline, the thoughtful version of it is no longer available. And when the spend is weighed, weigh it against what the current environment actually costs to run today, including the headcount, the workarounds, and the business not written. Not against zero.

Addressing it does not mean replacing everything. Start where the operational friction is highest, fix the underlying data or process problem, prove the model, and build outward from there.

That work usually means one of these: a golden source for reference data; a firm-wide data warehouse; portfolio return and attribution databases; replacing spreadsheet-driven processes; selecting and implementing new platforms; or improving the controls and versioning around what already exists.

My work is those problems, across investment operations, data, trading, risk, treasury and reporting. The objective is not modernization for its own sake. It is an operating environment that can support the business you are trying to build.

Written on this: Buy the Exit, Not the Demo · Git for the Middle Office · Nobody Priced the Environment You Already Have

Model: Cost of Current State Worksheet

Investment data you can trace back to source

A number in a report you published three years ago should still be there when somebody pulls it again. Under the standard price adjustment convention it is not. A price quoted at $48.56 in a 2016 report reads $37.26 this morning. Nothing was restated, no vendor issued a correction, and no reconciliation caught it, because there was never a break to find.

The same thing happens quietly in curve construction. Six interpolation methods can reproduce every quoted node on a SOFR curve exactly, so two systems using different methods reconcile clean. Ask them for the rate between the quotes and they span three basis points. Ask them for the forward and they span fifty.

Neither of these is a data quality problem. Both are conventions nobody wrote down, and they decide whether your history is reproducible and whether two systems that agree actually agree. I find which convention your system is on, show what it costs, and get it written down as a decision rather than an inheritance.

Written on this: The History You Published Last Year Has Changed · Interpolation Is a Policy Choice Nobody Wrote Down

Models: Forward vs Backward Price Adjustment · Curve Interpolation Comparison

New asset classes the platform was never built to hold

When a strategy gets wrapped to reach a different investor, the wrapper is an operations decision before it is a distribution decision. The structure sets the pricing frequency, the liquidity mechanics, the reporting obligations and the reference data somebody maintains every day afterward. The investment case gets debated for months. The operational consequences of the wrapper often get one meeting.

The same gap shows up in onboarding generally. Adding an instrument type is made as a trading decision and arrives as an engineering project, because the terms live in credit agreements and servicing notices rather than in a feed.

I work that problem from the instrument outward: model it correctly, establish the reference data, then confirm that everything downstream can process what you intend to own. Including the case where the honest answer is that the platform cannot carry it.

Written on this: The $500 Minimum Is an Operations Decision

Derivatives Operations and Governance

The specialty underneath it. Hedging programs, effectiveness measurement, and documents built to be tested rather than filed.

Governance documents that are control structures

A derivative use plan that lists permitted instruments tells you nothing. A plan that names who decides, under what limit, against which liability, and what evidence is retained is a control structure, and it is the one an examiner can actually test. The same holds for the section nobody wants to write: the risks you are deliberately not hedging, and why.

I write the document, the limit structure underneath it, and the reporting that makes it auditable.

Written on this: A Derivative Use Plan Is Not a Permissions List · Write Down What You Are Not Hedging · The Hedge You Can Actually Run

Hedge effectiveness that can see sign and persistence

Three conventional effectiveness measures will return three different verdicts on the same book. A hedge can sit inside the 80% to 125% band on every single day of a quarter while running one-sided on 271 days out of 362, because an absolute band cannot see sign and cannot see persistence. And the sizing convention decides which test you appear to pass before a single hedge is placed.

I review how effectiveness is measured today, build the report that shows what the band conceals, and get the sizing convention written down as a decision rather than a habit.

Written on this: Your Hedge Is Worth More and Does Less · Percent of Max Is a Fact About Your Entry Price · The Greeks Are an Operating Budget

Models: Cap Ladder Hedge Analyzer · Call Spread Exit Calculator · Option Greeks Calculator

Derivatives operations and documentation

Carriers end up bilateral by drift. A dealer proposed the structure, or that is how the last one was done, and six months later nobody can reconstruct whether it was a decision or an accident. In the file, the two look identical. The exemption most carriers believe they are relying on does not cover insurers at all.

I write the rationale per trade: the index, the convention, the class it falls outside, and the liability the terms were built to match, along with the valuation that supports the mark.

Written on this: The Clearing Exemption You Don't Have

Model: Fixed vs Float Swap Model

How the work runs

One person does the work, and it is the same person who wrote the document. Most people who can build the system have never written a derivative use plan. Most people who write derivative use plans have never built anything. Thirty years across the front, middle and back office is what puts both on the same desk.

Engagements are scoped to a named deliverable: a document, a model, a report, a database. Not an open-ended retainer.

If you have a middle office running on spreadsheets you no longer trust, or a derivatives book and a regulator asking questions, that is the conversation to have.