What is Decision Latency?

A well-run services firm can quote its delivery lead times, its payment terms and its time to hire almost to the day. Ask the same firm how long it takes to make a decision once the evidence is in, and the room goes quiet. The interval between we knew and we chose is one of the most expensive unmeasured intervals in business.

That interval is decision latency. The business intelligence researcher Richard Hackathorn decomposed the delay between a business event and a useful response into three parts: data latency, the time for the fact to arrive; analysis latency, the time to understand it; and decision latency, the time to choose what to do about it. His point was that the value of acting decays while the clock runs. Two decades of engineering have collapsed the first two intervals: facts now arrive in near real time and dashboards render them in seconds. The third is governed not by technology but by calendars, ownership and nerve, and it has barely moved.

Decision latency is the time from signal to decision, and it is the silent tax on every option with an expiry date.

Where the time hides

Decision latency rarely looks like waiting. It hides in three respectable places:

• Waiting for the meeting. Signals arrive continuously; forums convene monthly or quarterly. A decision-ready signal that lands the day after the review waits a full cycle for its hearing, and nobody experiences this as delay because the calendar is being followed perfectly. The cadence of governance quietly becomes the speed limit of the business.
• Waiting for the owner. When decision rights are unclear, the decision queues while the organisation works out whose it is. The question sits between functions, each sincerely believing another holds the pen, and this is the slowest wait of all because nobody knows the clock is running.
• Waiting for courage. The analysis is finished and the answer is visible, but the downside has an owner and nobody volunteers. So more analysis is commissioned, not because the decision needs it but because requesting data is the polite form of not deciding. The request for one more cut of the numbers is often deferral wearing diligence as a disguise.

The silent tax

Latency never appears on a profit and loss statement under its own name. It books its costs as other things: contractor premiums, expedite fees, discounts given late in a negotiation, renewals rescued at a price. The mechanism is always the same. Options have expiry dates, and on a fast-moving constraint the option list degrades while the decision waits: the cheap option needs lead time, the moderate option needs a booking, and the expensive option is always available. Every significant decision has a decision window, and latency is the fraction of that window spent not deciding.

The cruellest property of latency is that the decision finally made is often identical to the one available on day one. Nothing was gained by the wait; only the price changed. And deferral compounds: each unmade decision constrains the ones behind it, accumulating into decision debt, the backlog of choices the organisation is paying interest on without ever having made them.

One concrete example

Clearly illustrative, with no customer implied. A services firm sees a client’s volumes rising early in a quarter: the signal is clean, measured and sitting in the weekly numbers. The cheap response is to hire and train ahead of the ramp, an option that needs several weeks of lead time. But capacity is reviewed quarterly, so the signal waits for the meeting. At the meeting it emerges that nobody is sure whether the account lead or the delivery director owns a hire-ahead call, so the question is taken away. By the time it comes back answered, the training option has expired, and the ramp is staffed with contractors at a premium, under pressure, with the client watching. The ledger will record a delivery cost overrun. It was nothing of the kind. It was a latency cost: the decision made in week eleven had been available in week two, at a fraction of the price.

Deciding on evidence time, not calendar time

Latency shrinks when the machinery of deciding is rebuilt around evidence rather than around the calendar, which is the heart of decision intelligence. Facts land as versioned evidence, and a changed fact triggers re-evaluation the day it changes rather than at the next scheduled review. The window becomes an object: a decision window with an owner and an expiry, so an ageing decision emits a signal instead of sitting silently between functions. And the cost of waiting becomes visible, because every candidate option carries the earliest date it clears every constraint that binds it, and the worst constraint decides: when delay pushes an option past its expiry, the ranking visibly reorders, and the room can watch the option list worsen in front of it. A person still makes the call. The difference is that they make it while the good options are still on the table.

Speed is not the goal; a reckless decision made instantly is no achievement. The goal is to spend the window on judgement instead of on queueing. Measure the interval between signal and decision, and the tax starts shrinking the moment it is seen.