I build systems for wealth and investment businesses: portfolio accounting, holdings and transaction reconciliation, fee computation, client reporting and risk analytics — for advisors, family offices and portfolio managers in India and the UK.
What hides behind wealth management’s PDFs?
Operational archaeology, mostly. The statements look institutional; the process behind them is holdings reconciled quarterly by hand, corporate actions applied when noticed, fees computed in a spreadsheet only one person understands, and review packs assembled by copy-paste the night before the meeting. It survives on trust — until the day a client’s CA asks why the XIRR in the statement disagrees with their own computation, and “the spreadsheet had a stale price” becomes the answer that loses the AUM.
The failure is identical in Mumbai and Edinburgh; only the accent differs. And it’s not a diligence failure by the advisor — it’s what happens when a business whose product is accuracy about money runs its own operations on tools with no memory. Every number in wealth management is a claim against history: what was bought, when, at what price, with what fees, through what corporate actions. Tools that store balances instead of history can’t defend claims — they can only repeat them louder.
What does a wealth build include?
- The portfolio accounting core. Every buy, sell, dividend, split, bonus, fee and expense an immutable event; positions, cash and returns computed as projections over that history. It’s the ledger architecture wearing an investment jacket — and it means any number on any statement can be replayed from first principles, which is the entire game.
- Reconciliation as a daily habit. Custodian statements, broker contract notes and registrar feeds matched against internal truth automatically; breaks queued with reasons. Quarterly manual reconciliation becomes a daily automated check — the method is the same one that runs payment stacks, pointed at securities.
- Corporate actions handled properly. Splits, bonuses, mergers, rights — applied as events with effective dates, so historical returns don’t silently rewrite themselves and the pre-action lots keep their identity for tax lots and audit both.
- The fee engine, defensible. Slabs, hurdles, high-water marks, pro-rating, GST or UK VAT — computed, shown and reproducible. Fee disputes end when the computation is a report rather than a spreadsheet only Priya understands, and Priya gets her evenings back.
- Client reporting from one source. Statements, review packs, performance summaries generated from the ledger — numbers that tie across every page because they come from one place. The night-before assembly ritual retires.
- Risk views that earn their place. Allocation drift against mandates, exposure concentrations, and parametric VaR where mandates demand it — the portfolio system and risk modelling engagements built this at production grade: event-sourced truth underneath, Monte Carlo where the questions need it.
Why does transaction-truth beat snapshot tools?
Because wealth’s hard questions are all historical. What did this client actually earn, net of fees, since inception? What was the position before the merger? Why does this year’s opening balance differ from last year’s closing? Snapshot tools — aggregators, consumer portfolio apps, the custodian’s own portal — answer with today’s picture and a shrug about the past. Transaction-true accounting answers with arithmetic, because the past is the data model. That’s also what separates the productised version from the app-store aggregators: they collect; this accounts. An audit, a SEBI query, an FCA file review or just a sharp client’s CA — all of them are history exams, and the system that stores history passes them by construction.
What changes in the first quarter?
Month one: the daily reconciliation stands up, and the breaks it surfaces — the missed dividend, the wrongly-applied bonus, the stale price — get fixed while they’re small. Month two: fee computation moves into the engine, and the quarter’s invoices go out with computations attached; nobody disputes arithmetic they can read. Month three: review season runs from generated packs — assembly time falls from a week to a button, and the meeting starts at insight instead of at verification. Clients notice the calm before they notice anything else; calm, in this business, is the product. Chuffed clients, sorted operations.
Related reading
/work/portfolio-system — the build in production · /work/risk-modelling — the quant layer · /products/portfolio-manager — the productised version · /blog/the-t-1-problem — settlement truth, explained
The review-week test
If your quarterly reporting week is genuinely a week — /contact. It should be a button, and the first custodian reconciliation usually proves the point inside a fortnight.
Questions I actually get
Do you handle SEBI and FCA compliance requirements?
The systems evidence, yes — RIA and PMS record-keeping discipline in India, FCA-side records for UK operations: transaction trails, suitability documentation hooks, fee transparency, access logs. Filings and regulatory interpretation stay with your compliance owner; the build makes their evidence a query instead of a quarter-end assembly.
Which custodians and brokers can you reconcile against?
File and API both — custodian statements, broker contract notes, registrar feeds. Normalising a new counterparty's format is routine work, not a project: the reconciliation engine treats formats as dialects of the same comparison, which is why adding your third custodian costs days rather than months.
Is a family office too small or too odd for this?
Family offices are the ideal fit — complex enough to need transaction-true accounting across entities, trusts and individual books, small enough that the institutional platforms ignore or overcharge them. Multi-entity consolidation with clean separation is the core design, not an enterprise add-on.
How do you handle unlisted and alternative investments?
As manually-marked positions with the same accounting rigour: acquisition events, valuation marks with dates and sources, distributions as entries. The XIRR on your AIF or unlisted position becomes as defensible as your listed book — the marks are opinions, but the arithmetic around them is fact.
What does the VaR and risk layer add for an advisory practice?
Drift against mandates caught weekly instead of at review time, concentration named before clients ask, and — where mandates want it — parametric VaR computed from current positions rather than last quarter's snapshot. The risk-modelling case study shows the engineering; the practice benefit is conversations that start from numbers.
Can clients see their own portal?
Yes — statements, holdings and performance served from the same single source the advisor sees, so the client's screen and your screen cannot disagree. Transparency built on one truth is a retention feature: the awkward reconciliation conversation stops existing because there is nothing to reconcile.