Learning objective
By the end of this lesson, you can identify a pivotal assumption in a weighted matrix and use sensitivity analysis before presenting the result as robust.
Why it matters
A weighted total may change when a credible score, weight or normalisation rule changes. The arithmetic can be correct while the recommendation remains unstable.
Core concept: a rank is conditional
A Decision Matrix ranking is conditional on:
- which alternatives are included;
- which criteria are separate;
- how scales are defined and normalised;
- which weights represent preferences;
- what evidence supports each score.
Sensitivity analysis changes plausible inputs and records whether the leading alternative changes.
Visual explanation
Text alternative: start from the base ranking, identify pivotal weights and scores, vary them across credible ranges and observe whether the winner changes. Stability supports a more robust recommendation; reversal requires a conditional choice or more evidence.
Worked example
Platform A leads B by 1.2 points. A's accessibility score is based on a sales demonstration and carries 20% of the total weight. When that score varies from 5 to 9, B leads below 7.
The decision record should state: “A is preferred only if the independent accessibility test scores at least 7 on the defined scale.” The model has turned an argument into an evidence condition.
Common mistake
Mistake: varying every number randomly.
Use credible ranges tied to evidence or stakeholder preference. Start with high-weight, low-confidence cells and criteria whose definitions are contested.
Quick check
A one-point change in a low-confidence score reverses the top two options. What should the team do?
A. Hide the sensitivity.
B. Add decimal places.
C. Obtain evidence or make the decision conditional on that assumption.
D. Increase the winner's weight.
Answer: C. The pivotal uncertainty should govern the next evidence step.
Practical prompt
Take one matrix. Identify its two highest-weight, lowest-confidence cells. Define credible ranges and state what decision condition follows if the rank reverses.
Summary
- A weighted rank is conditional on model choices.
- Sensitivity tests credible uncertainty.
- Focus on pivotal, low-confidence inputs.
- An unstable rank becomes an evidence or governance condition.
Next lesson
Return to the Better Decision Making path and compare the matrix with Kepner–Tregoe and Pairwise Comparison.