Architecture
A client asks what they made this year. In most firms there are four places to get the answer, and no one is responsible for the difference between them.
It is the simplest question in the business, and in most firms there are four places to get the answer. The client portal. The quarterly report. The proposal software the advisor used to win the relationship. The planning tool that projects it forward.
Pull all four. In most stacks they will not agree to the basis point. In a fair number they will not agree to the percentage point. And the advisor, standing in front of the client, has to pick one and hope the client never sees another.
That is what makes this hard to fix and easy to ignore. None of those systems is broken. Each one is computing a return correctly, using a convention it is entitled to use.
One weights by time and one weights by dollars, and the client who added money in March gets two different truths. One counts accrued interest and one does not. One treats a transfer in as a flow and one treats it as a starting balance. One took its snapshot Friday and one took it at month end. One annualises a partial year and one refuses to.
Any of those choices can be justified in a room full of accountants. The problem is not that a system chose wrong. The problem is that four systems each chose, independently, and nobody in the firm is responsible for the difference.
Every platform in this category advertises integration, and most of them deliver it. Data moves. The planning tool sees the holdings. The proposal tool sees the accounts. The portal shows the balance.
What almost never moves is authority. Both systems now hold the data, and both are still willing to compute a figure from it. Integration made the inputs agree. It did not decide who is allowed to be right about the output.
That distinction sounds academic until a client, a prospect, or an examiner puts two documents side by side and asks a question you cannot answer in one sentence.
For every number a firm puts in front of a client, exactly one system computes it. Everything else displays it.
Displays, not syncs. A figure that is synced between two systems is a figure that was computed twice and then argued about, and the argument is settled by whichever job ran last. Displaying means the proposal, the portal, the quarterly package and the review deck are all reading the same computation, produced once, from the book of record, with a date attached to it.
The rule has a second half that matters as much. A system that owns a number is not permitted to invent numbers it does not own. The side of our platform that holds the suitability record does not compute portfolio risk. The side that computes portfolio risk does not decide what a client is suitable for. Neither one is allowed to fill in the other's blank because a screen needed something to show.
This is not a policy you can adopt later. It is a consequence of where the calculation lives.
If four products each shipped with their own performance engine — and they did, because each was sold standalone before it was sold to you as part of a suite — then the firm that assembles them inherits four engines and no referee. You can write a memo declaring one of them authoritative. The other three will keep computing anyway, because that is what they were built to do, and their answer is what your client sees on the screen in front of them.
The only durable version of the rule is architectural: there is one place the calculation happens, and the other surfaces have no engine to fall back on.
Pick one household. A real one, ideally one with a mid-year contribution and more than one account.
Write down year-to-date return as it appears on every surface a client could plausibly see. The portal. The last quarterly. The proposal, if it is recent. The planning file. The performance report your ops person runs.
Then ask two questions. Do they match? And if they do not, which one is the firm prepared to defend in writing?
Most firms have never run this. The ones that do run it are usually surprised, and the surprise is rarely at the size of the gap. It is at the discovery that nobody in the building can say which number is supposed to win.
We are not going to publish our return conventions, the order the calculations run in, or how the book is reconciled before any of it is computed. Those are the part that took years.
The rule itself is worth publishing, because it is not a secret and it is not hard to evaluate. Ask any platform you are considering which system owns each number on a client statement. If the answer takes more than a sentence per number, you have found the thing that will cost you an afternoon a quarter for as long as you own it.
Questions this did not answer? Ask them directly — that is what the twenty minutes is for.
Book 20 minutes with Kyle