Loading
Case study
The data pipeline and portfolio analytics behind a wealth-management firm's investment process, from monthly attribution to the diffusion-index risk indicator that guides its managed-account trading.
This is the investment-analytics half of my consulting work: a long-running engagement with a wealth-management firm and the in-house macro-research service run under the same leadership, split roughly half-time between the two. Both are anonymized here, and everything below is my own contribution.
It is the part of my record that pairs most naturally with the compliance and supervision decade: the same close read on a portfolio and the data underneath it, from the operating side rather than the audit side.
A small RIA has the same portfolio-analytics needs as a large one on a fraction of the infrastructure: attribution has to come out on time every month, model portfolios have to stay in policy through drift and cash flows, and client meetings have to be prepared without an advisor rebuilding the same deck by hand.
The firm also manages risk in its accounts with a data-driven, rules-based process rather than gut feel, and that process is only as good as the data under it. It runs on custom diffusion indexes that blend a wide set of macro indicators onto common statistical scales into a single published risk signal: when the composite runs too hot or too cold, that is the cue to take risk down or add it back. Keeping that signal honest is a data problem before it is a research one, and it was being maintained by hand in spreadsheets.
On the wealth-management side, I built the monthly analytics stack: portfolio-weight and drift monitoring across model-portfolio households, trade-rebalancing reports, monthly performance attribution, and the model-portfolio fact sheets and holdings summaries advisors take into client meetings. I wired the Redtail CRM API into a parameterized report pipeline so client-specific content (holdings, allocations, review-meeting talking points) is populated from the system of record rather than assembled by hand.
On the risk side, I maintained the firm's diffusion-index risk indicator and reverse-engineered it from a manual Excel workbook into an automated R script. I tracked down every underlying value across the macro domains it watches (recession and labor signals, energy and oil, monetary policy, and the equity market), computed the summary statistics that put each series on its scale, and turned the monthly recompute from a spreadsheet grind into a repeatable run.
On the broader research side my role was production rather than authorship: the principals wrote the commentary, and I sourced the data, turned useful topics into charts, kept the standing charts current across releases, proofed the finished document, and handled distribution to the firm's network of financial-advisor subscribers.
The parts worth pointing at
Where the engineering decisions actually mattered.
Client-branded artifacts stay with the client; happy to walk through the architecture of the attribution pipeline, the Redtail integration, or the way the monthly index publications are produced.