Decision latency is the time between a change appearing in your financial data and your organization making a decision about it. It covers three delays: getting the data, explaining why the number moved, and agreeing what to do. For most FP&A teams it is measured in days or weeks, not hours.
Why decision latency matters
A variance is only useful while you can still act on it. If a spend overrun takes two weeks to explain, the budget for that period is already committed, the business has already formed its own story about the miss, and the next review treats it as history. The number was right. It just arrived too late to change anything.
That is why decision latency is a better measure of a finance function than close speed alone. A team can close the books on day five and still take another week to explain what the books say.
The three parts of decision latency
| Part | Starts | Ends | Typical causes of delay |
|---|---|---|---|
| Data latency | The event happens (a deal closes, an invoice posts) | The data is available and reconciled | Batch exports, close timing, systems that disagree |
| Explanation latency | The number is available | Someone can say why it moved, with evidence | Manual tracing across the ledger, CRM and spreadsheets; the context lives in one person's head |
| Agreement latency | The explanation exists | A decision is made and recorded | Waiting for the next meeting; competing versions of the number; no shared record of the reasoning |
Most investment goes into the first part (faster close, better pipelines). In our conversations with finance leaders, the second part is where most of the time goes: the explanation is rebuilt by hand every time the question comes up.
How to measure decision latency
Pick one recurring decision, such as approving a reforecast after a material variance. For the last few cycles, record four timestamps:
- T0: the date the underlying event happened.
- T1: the date the reconciled number was available to finance.
- T2: the date finance could explain the variance with evidence.
- T3: the date the decision was made and written down.
Decision latency is T3 minus T0. The split tells you where to work: T1 minus T0 is data latency, T2 minus T1 is explanation latency, T3 minus T2 is agreement latency.
A software cost line comes in over budget. The invoices post during the month (T0, day 0). The books close on day 5 (T1). Finance traces the overrun to an annual renewal billed early on day 11 (T2). The reforecast is approved at the monthly review on day 16 (T3). Decision latency is 16 days, and 6 of them were spent explaining the number.
What causes high decision latency
High decision latency is a structural problem, not a tooling or talent problem. The common causes:
- Disconnected systems. The ledger, the CRM, the billing system and the warehouse each hold part of the answer, and they disagree on timing and definitions.
- Explanations that live in people's heads. Once a variance is resolved, the reasoning is not recorded, so the next occurrence starts from scratch.
- Verification work. Most of the time goes into checking that numbers are right, not into calculating them.
- Meeting-paced decisions. Even a ready explanation waits for the next review on the calendar.
Decision latency vs. related metrics
| Metric | What it measures | What it misses |
|---|---|---|
| Days to close | How fast the books are final | Everything after the close: explanation and decision |
| Data latency | How fresh the data is | Whether anyone can explain it |
| Forecast accuracy | How close the plan was to actuals | How fast the team reacts when it is wrong |
| Decision latency | Time from event to recorded decision | Decision quality, which needs its own review |
How to reduce decision latency
- Measure it on one recurring decision for three cycles before changing anything.
- Reconcile once, at the source. Agree how ledger, CRM and billing records match, and keep that logic versioned rather than rebuilt in each spreadsheet.
- Make every number traceable. Anyone should be able to click from a reported figure to the transactions behind it.
- Record the explanation and the decision. Keep the root cause, the owner and the approval with the number, so the next cycle starts from the last answer.
- Decide between meetings. When the explanation and evidence travel together, approvals do not need to wait for the review.
Together these are what we call decision infrastructure.
Frequently asked questions
Is decision latency the same as close cycle time?
No. Close cycle time ends when the books are final. Decision latency ends when a decision is made, which is usually days or weeks later.
What is a good decision latency for an FP&A team?
There is no published benchmark. Start with your own baseline on one recurring decision, then aim to cut explanation latency first, since it is usually the largest share.
Does AI reduce decision latency?
Only if the output can be trusted without being re-checked by hand. If every AI answer has to be verified manually, the explanation step does not get shorter. Traceability to source data matters more than speed of generation.