Decision quality metrics: how to measure the decision, not the outcome
The essentials in 30 seconds
- Decision quality metrics score the process, the estimates and the follow-through of a decision, not its outcome.
- An illustrative quarter of 20 decisions gives a process score of 4.1 out of 5 and 80 percent of reviews completed.
- The range hit rate is 69 percent against a stated confidence of 70 percent.
- 3 of 20 decisions were reversed within 90 days, a rate of 15 percent.
- The EQOS Asia Sentinel service recomputes a decision model monthly (EQOS Asia method page, as of October 2026).
Decision quality metrics measure the decision as made, not the outcome that followed. A scorecard of 6 metrics, built on a decision journal, shows whether the process, the estimates and the follow-through of a team are improving. This page extends the guide to keeping a journal and works through an illustrative quarter of 20 decisions as of October 2026.
On this page
Compare the EQOS Asia offers and request a 30-minute pre-diagnostic.
What are decision quality metrics and why measure the decision itself?
Decision quality metrics are numbers that describe how a decision was made and followed through. They stay separate from profit or growth, which also reflect luck and timing. A team that measures only outcomes learns slowly, because 1 good result can hide a weak process and 1 bad result can hide a sound one.
A decision journal supplies the raw material. Each entry holds the estimate, the confidence, the process checks and the review. Counting these fields across 20 entries in a quarter produces a scorecard with 6 metrics. The scorecard fits on a page and is reviewed in 30 minutes.
Which process metrics belong in a decision quality scorecard?
Three metrics describe the process. The average process score takes the checks of each entry, such as 3 options considered and a dissenting view heard, and averages them out of 5; the illustrative quarter scores 4.1. Review completion counts the entries reviewed on their due date: 16 of 20, or 80 percent.
The third metric is decision latency, the days between the question and the decision. The illustrative median is 9 days. Latency informs in both directions: a very long median signals a stuck process, and a very short one on costly choices calls for a second look at how much evidence was gathered.
Which calibration metrics show whether decision estimates can be trusted?
Two metrics test the estimates. The range hit rate is the share of reviewed entries whose actual value fell inside the written range. In the quarter, 11 of the 16 reviewed entries landed inside, or 69 percent. The entries stated a confidence of 70 percent on average, so the calibration gap is 1 point.
The second metric is the average absolute estimate error, the distance between estimate and actual as a share of the actual. The illustrative average is 22 percent. A small gap and an error that falls from quarter to quarter indicate that the numbers of the team can be used in the next go decision at face value.
How does the reversal rate show the quality of a decision?
The reversal rate counts decisions undone or materially changed within 90 days. In the illustrative quarter 3 of 20 were reversed, a rate of 15 percent. A reversal is a signal about the inputs: a missing stakeholder, an assumption nobody checked or pressure to decide before the evidence arrived.
Read each reversal with the journal. If the entry shows a weak process score, the fix lies in the routine. If the entry shows a strong score, new information arrived after the choice, and the reversal reflects a healthy response. Separating the two readings keeps the metric useful and keeps people open about changing course. A short note beside each reversal, with the date and the new fact, completes the record for the quarterly review.
What does a quarterly decision quality scorecard look like?
The scorecard lists 6 metrics with the value, the target and the status. Process score 4.1 against a target of 4.0 is on track. Review completion of 80 percent against a target of 90 needs attention. Latency of 9 days against a ceiling of 14 is on track. Each line links back to the journal entries behind it, so a reader can open the evidence for any number in a click or two.
Range hit rate of 69 percent against a stated 70 is on track. Estimate error of 22 percent against a target of 20 needs attention. Reversal rate of 15 percent against a target of 10 needs attention. 3 of 6 metrics are on track, which tells the team to fix review follow-up first.
How do you set targets that keep decision quality metrics honest?
Set targets as ranges around current values and move them 1 step at a time. A target of 4.0 for the process score is realistic when the quarter shows 4.1, and a jump to 5.0 invites box ticking. Report the metrics for the team as a whole and keep individual rankings out of the scorecard, so that people record entries candidly.
Protect the data with a light audit. Each quarter, one reviewer reads 3 random entries and compares the written process checks with the evidence in the file. Findings go into the next template version. Rotating the reviewer builds shared standards and keeps the scorecard tied to the real quality of the decisions.
How does the EQOS Asia decision engine support decision quality metrics?
The Sentinel service lets a model be reused monthly and recomputed on updated inputs, with real-time alerts when a threshold is approached (EQOS Asia method page, as of October 2026). A monthly score gives the scorecard a regular data point beside the review metrics above.
The Audit pillar scores a system on 9 structural dimensions and lists 3 priority risks (EQOS Asia method page, as of October 2026). A reader can try a first score in the simulator, and the 30-minute pre-diagnostic, scheduled in Singapore time (SGT), applies the equation to a real decision.
See also in this guide
Frequently asked questions
What are decision quality metrics?
Decision quality metrics are numbers that describe how a decision was made and followed through, such as the average process score, review completion, decision latency, range hit rate, estimate error and reversal rate. They stay separate from outcomes, which also reflect luck and timing.
How many decisions do you need for a scorecard?
Start with the 20 decisions of one quarter. The numbers show tendencies and leave room for chance, so treat early readings as hypotheses and confirm them over 2 or 3 quarters. A team with fewer decisions can score the biggest 10 each quarter and keep the template unchanged.
What is a good range hit rate?
A good range hit rate matches the stated confidence. In the illustrative quarter the entries stated 70 percent on average and 69 percent landed inside the range, a gap of 1 point. A hit rate far below the stated confidence signals ranges that are too narrow for the uncertainty involved.
How can the EQOS Asia pre-diagnostic help with decision quality?
A free 30-minute pre-diagnostic, scheduled in Singapore time (SGT) and held under NDA, applies the EQOS Asia equation to a real decision (EQOS Asia contact page, as of October 2026). The result gives the scorecard a computed starting point to track from month to month.
Test your knowledge
1. What is the average process score in the illustrative quarter?
- 4.1
- 3.5
- 5.0
Show the answer
4.1.
2. How many of the 20 decisions were reversed within 90 days?
- 1
- 3
- 6
Show the answer
3.
3. What is the median decision latency?
- 14 days
- 5 days
- 9 days
Show the answer
9 days.
4. What is the calibration gap in the example?
- 10 points
- 1 point
- 5 points
Show the answer
1 point.
5. Which metric does the scorecard fix first?
- Review completion
- Latency
- Range hit rate
Show the answer
Review completion.
About the author
Stéphane Jambu
President, EQOS Forecast Asia
Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile
Sources: EQOS Asia method page (eqos-asia.com/method.html), simulator page (eqos-asia.com/simulator.html), and contact page (eqos-asia.com/contact.html), consulted on 22 October 2026. Page updated on 22 October 2026.