Trust Accounting Software Problems: The 7 That Cost Real Money, and How to Avoid Them

Why does the trust document reading get left out of scope so often?

A proposal lists a dual ledger for principal and income as a feature. It is not a feature. It is a rules engine whose rules live in several hundred separate legal instruments, most of them drafted by different lawyers across fifty years, each of which overrides the statutory default in some respect. The Uniform Principal and Income Act and its successor the Uniform Fiduciary Income and Principal Act give you the framework. The deed gives you the answer.

This is specific to fiduciary work. In almost every other build, the business rules can be extracted from a handful of people in two workshops. Here they are extracted by a qualified reader working through documents at a rate of a few per day, recording the interests, the distribution standard, the allocation treatment for the receipts that actually arrive, and the article each of those came from. That work is expensive, it is nobody's favourite line item, and it is the first thing cut when a fixed price needs to come down.

The fix: price the encoding as its own staffed workstream with a named reviewer who is qualified to read the documents. During discovery, sample thirty instruments and measure how many depart from the statutory default. That measured rate, not an assumption, is what you extrapolate across the book. Then ship with an explicit unverified state on every account whose rules have not been read, so the gap is visible on a report rather than hidden behind a plausible default.

What goes wrong with migrating decades of allocations and cost basis?

Conversion is where trust builds actually fail. The legacy platform hands over an income balance, a principal balance, cost basis by lot and a fee history, and every one of those is a number with no derivation attached. The income balance is not a bookkeeping figure. It is a legal position between an income beneficiary and a remainderman with opposed interests, and if the old system carried it wrongly you inherit the error under your own name the moment you rely on it.

The history makes it worse. Accounts have changed situs, been through two custodian conversions, and passed through a platform that no longer exists. Cost basis on assets transferred in decades ago arrives blank or carries a figure nobody can source. Fee exceptions granted verbally in 2011 explain differences the data cannot.

The fix: treat it as a reconciliation and review programme, not a data load, and budget it separately. Reconstruct and review the opening principal and income split rather than accepting it. Flag every account where the split cannot be derived and route it to an officer before the system is relied on. Run parallel for at least one full statement cycle with a period end tie out. Prioritise accounts with an imminent court accounting or an active beneficiary dispute for early review, because those are the ones where an inherited error is discovered publicly.

Why do the custodian and market data feeds break after launch?

The feed covers marketable securities. Your book does not. Real property, closely held interests, notes receivable, mineral rights and tangible personal property never appear on a custody file, so on day one a share of the trust value is being carried by hand in a system that was tested against clean feed data.

The second break is subtler and costs money. Trustee fees are computed on market value at a valuation date. Prices get restated, corporate actions arrive late, and a distribution posts after the fee run. So a fee that has already been charged, taken against principal or income under the document, and reported to a beneficiary, is now based on a valuation the custodian has since changed. Systems that recompute on read will silently produce a different number the next time anyone looks, which means your statement and your screen disagree and neither is wrong.

The fix: treat every inbound price and position as evidence with an as at stamp rather than as current truth. Hold the fee run until the valuation for that date is confirmed. Handle a restatement as an adjusting entry in the open period with a reference to the original, never as an overwrite. Carry specialty assets as first class holdings with their own valuation cadence, the basis for each valuation, and their own income and expense treatment recorded once and applied consistently.

What happens when discretionary distribution evidence is not covered?

Institutions build the ledger and leave the committee where it was, which is email and a Word memo on a shared drive. The ledger then shows a payment with a memo line, and the reasoning that justified it lives in an inbox belonging to an officer who has since retired.

A discretionary distribution is the highest liability event in trust administration, because the trustee exercised judgement and judgement is what gets challenged. Years later a beneficiary asks why one sibling was funded and another was not. The defence is documentation created at the time: the request, the standard applied from the instrument, the beneficiary information considered, the committee discussion, the decision and any dissent. Recollection is not a defence, and neither is a payment with a memo line.

The fix: make the distribution a structured case that cannot be approved without its required fields, routes to the correct committee by amount and account type, records each member's position, and links the resulting memo immutably to the ledger entry that moved the money. Give recurring distributions the same treatment, because a monthly payment set up years ago under circumstances that have since changed is exactly the item counsel will open with. Institutions that do this usually find the committee meeting gets shorter, since the preparation is already structured.

Should you build custom or configure what you already own?

If you administer a few hundred accounts, mostly revocable, invested largely in marketable securities, across one or two states, do not build. Accutech Cheetah and InnoTrust are built for exactly that shape of book and the economics are not close. A first custom build will not match the fiduciary detail those products have accumulated, and you would be paying to reproduce it.

If you are a bank trust department already running FIS Global Plus or SEI Trust 3000, the question is almost never whether to replace the accounting engine. Those platforms encode decades of principal and income correctness that is genuinely expensive to reproduce, and replacing them puts your ledger at risk to fix problems that are not in the ledger. The right question is what to build alongside them.

Build, or build around, when two or more of these hold. You administer across several states with different principal and income statutes and different court accounting formats. A material share of your book holds real property or closely held business interests your platform treats as a memo entry. Your fee exceptions are tracked outside the system and cannot be reconciled. Your distribution evidence lives in email. Or you have been through an examination where producing the historical record took weeks.

How do hidden costs get into the quote?

Five items reliably arrive after signature in this category. Jurisdiction count, because each additional state brings its own principal and income statute and its own court accounting schedules, and a quote written against one state does not scale linearly to four. Specialty assets, since each class needs valuation, expense and income handling that clean feed data never exercises. Migration, which is a reconciliation programme staffed by your officers rather than a developer task, and which routinely costs more in your people's time than in the invoice. Court supervised accounts, which carry formal filing requirements the demo never touched. And the document encoding volume, which is a function of how many instruments depart from the default, a number nobody measured before pricing.

The fix: ask for the price broken out by jurisdiction and by asset class, then ask directly what the number becomes if a second state is added and if closely held interests come in scope. Ask who on your side is expected to staff the parallel run and for how many hours a week. A developer who cannot answer those without recalculating has priced a demo, not your book.

What separates a build that works from one that fails here?

Ask the team to explain principal and income allocation back to you before they propose anything. If they describe two ledger codes rather than two competing beneficial interests governed first by the instrument and then by state law, stop there. That single concept carries the whole category and it does not exist anywhere else in software.

Ask how a statement is stored. The issued artefact must be retained exactly as sent, not regenerated on demand from current logic and current data, because a regenerated statement is not evidence of what a beneficiary received. Ask how an account position as at a date twelve years ago is reconstructed. If the design updates rows in place, it cannot answer a dispute and should not be built. Ask how a closed accounting period is protected once a court has approved it, and expect the period to lock with later corrections posted as adjusting entries in the open period. Ask what they have migrated off a legacy fiduciary platform and listen for parallel running and a tie out rather than a conversion weekend.

Then settle ownership before kickoff. You should hold the repository, the infrastructure accounts, the documented schema and the right to hire anyone else. Trusts outlast software companies by decades, and a trustee who cannot read its own fiduciary records without a vendor's cooperation has created an exposure no service agreement covers.