Positions

When four advisors each answer a quarter of the question, someone has to reconcile them. That someone is usually the client, and it is the part of the work nobody was engaged to do.

A decision of any size touches strategy, tax, law and operations at once. The conventional arrangement sends each part to a different specialist, each of whom answers well within their own frame and is not accountable for the seams.

The seams are where the cost sits. A structure that is efficient for tax can be unfinanceable. A contract that allocates risk beautifully can make the operating plan impossible. None of the individual answers is wrong; the combination is.

Reconciling them requires holding all four frames at once and being willing to trade one against another — which is a judgement, not a summation. It is also work that no one advisor in the conventional arrangement has been asked to do, so it defaults to the client, who is the least equipped and the most distracted.

Holding the disciplines together is not a convenience. It is the difference between four correct answers and one usable one.

An advisor whose view reliably matches yours is an expensive way to hear your own voice — and the arrangement is comfortable enough that neither party notices.

The incentive runs one direction. Agreement is pleasant, renews the engagement, and is never obviously wrong at the time. Disagreement is uncomfortable, occasionally career-limiting for the person voicing it, and only vindicated later if at all.

So the drift is predictable. The advisor learns which conclusions are welcome and arrives at them more efficiently. The client experiences this as an advisor who understands the business well.

The correction is structural rather than personal. State the disconfirming evidence alongside every recommendation. Say what would have to be true for the advice to be wrong. Write it down at the time, so the record shows what was actually believed rather than what turned out to be convenient.

The test of an advisory relationship is not whether it is comfortable. It is whether the last three genuinely bad ideas were named as bad ideas before money was spent on them.

The figure on the page, the figure in the pack and the figure in the spreadsheet should be the same number because they come from the same calculation — not because someone checked.

Every reporting problem of consequence starts the same way: a figure is calculated once, then copied. The copy is used somewhere else. An input is corrected. The copy is not, and nothing announces that it has gone stale.

By the time the disagreement surfaces it is in front of a client, a lender or an auditor, and the question is no longer which number is right but why there were two.

The discipline is unglamorous and absolute. Store what was actually observed. Compute everything derived from it, every time, from one definition. Where a figure feeds a page, a pack and an export, all three call the same calculation rather than three implementations that agreed on the day they were written.

It is cheap to do at the outset and expensive to retrofit, which is why it is almost always retrofitted.