Sensitivity analysis: how far can one input move a decision score?
The essentials in 30 seconds
- A sensitivity analysis moves one input at a time and records the change in the score.
- Illustration with 9 equal weights: cutting one rating from 0.5 to 0.1 gives 0.418, and raising it to 0.9 gives 0.534.
- The EQOS Asia simulator offers 9 sliders from 0 to 100 and an influence cascade (EQOS Asia simulator page, as of October 2026).
- Inputs with wide swings deserve the most careful sourcing.
- A verdict close to the 0.618 line is the most exposed to a single input (EQOS Asia method page, as of October 2026).
A score is only as steady as its most sensitive input. A sensitivity analysis shows how far one input can move the result, and it tells you where to spend your effort on evidence. This guide explains the method, works through a simple example, and points to the tools EQOS Asia offers. It is one chapter of the mathematics of a decision engine.
On this page
Compare the EQOS Asia offers and request a 30-minute pre-diagnostic.
What does a sensitivity analysis measure in a decision score?
A sensitivity analysis measures how much the score changes when one input changes. The analyst picks an input, defines a plausible range for it, moves it across that range while the other inputs stay fixed, and records the resulting scores. The size of the swing is the sensitivity of the score to that input.
The method answers two practical questions. Which inputs matter most, so that they receive the most careful sourcing? And how close is the verdict to a threshold, so that a reader knows how much confidence it deserves? Both answers come from the same simple procedure. The procedure needs no special software, only a fixed equation and a patient log of results.
How is a sensitivity analysis run on one input at a time?
Five steps run the analysis. Pick the input to test. Set its low and high values from real evidence, not from convenience. Move it step by step while every other input stays fixed. Record the score at each step. Rank the inputs by the size of their swing. The ranking is the output: a short list that tells the team where evidence matters most.
Order matters in one respect. The base case must be written down first, with its sources, so every swing is measured from the same starting point. A reviewer can then repeat the analysis on any input and obtain the same figures, which keeps the exercise verifiable. The figure below shows the five steps.
Once the swings are ranked, a picture helps the committee read them. A tornado diagram draws one bar per input, with the widest swing at the top, so the inputs that deserve the most careful sourcing are visible in seconds. The same ranked list is the raw material for that chart, and it needs no extra data beyond the log of results.
Why can one weak dimension move a geometric score so far?
EQOS Asia states that its score aggregates the dimensions through a weighted geometric product, not an average (EQOS Asia WeWork back-test page, as of October 2026). The weights are proprietary, so this page uses a simplified model with 9 equal weights, only to show the shape of the effect. The values below are illustrations, not EQOS Asia results.
Start with all nine ratings at 0.5, which gives a score of 0.5. Raise one rating to 0.9, and the score rises to 0.534. Cut one rating to 0.1, the same distance in the other direction, and the score falls to 0.418. The fall is more than double the gain, because a low rating weighs heavily in a product. The table below sets the three cases side by side.
How close to the threshold can a single input push a verdict?
The EQOS Asia scale places the STOP zone below 0.618 (EQOS Asia method page, as of October 2026). Take a strong profile in the simplified model: eight dimensions at 0.7 and one at 0.2. The score is about 0.61, just under that line, although eight of the nine ratings look comfortable.
The reading is useful for planning. A single collapsed dimension can place a good-looking profile in the STOP zone, and a targeted fix on that dimension can bring the profile back across the line. A sensitivity analysis identifies that dimension before a committee meets. These figures remain illustrative and use equal weights.
What does the EQOS Asia simulator let you move in a decision score?
The EQOS Asia simulator turns the method into a hands-on test. Each of the 9 dimensions has a slider from 0 (collapsed) to 100 (perfect), and the live result updates as you drag. A Reset button returns every slider to 50, and a Random button draws values across the 9 dimensions to explore how the engine reacts (EQOS Asia simulator page, as of October 2026).
The Target button overlays the result of activating the top 3 levers, the 3 dimensions that weigh the most. Expert mode adds fine-tuning of weights and temporal modulation. Together, these controls let a reader run a sensitivity analysis by hand. The calibration coefficients stay proprietary and computed server-side (EQOS Asia simulator page, as of October 2026).
How do influence cascades change the reading of one input?
In a real system, inputs rarely move alone. The simulator shows an influence cascade: when a dimension moves, it drives the others, with first-order effects that are direct and second-order effects that pass through an intermediate dimension. The view is computed on a 9×9 matrix (EQOS Asia simulator page, as of October 2026).
A one-at-a-time analysis isolates a single input, and the cascade view shows how far its effect travels. Reading both gives a fuller picture: the isolated swing tells how sensitive the score is, and the cascade tells which neighbouring dimensions follow. The prediction view compares Now, Without action, and Top 3 levers over 6, 12, and 24 months (EQOS Asia simulator page, as of October 2026).
Inputs move together in real decisions, so a single swing tells only part of the story. A what-if analysis changes a small group of assumptions at once, gives the scenario a name such as supplier delay, and records the new score beside the base case. The result shows whether a plausible story pushes the verdict across the 0.618 line.
How do you use sensitivity results in a decision?
Use the ranking in three ways. Source the most sensitive inputs with the highest care, and accept rougher estimates for the others. Write down the range of each sensitive input next to the score, so readers see a band, not a point. Prepare a lever plan that targets the dimensions with the largest swings.
A 30-minute pre-diagnostic by video, scheduled in Singapore time (SGT), applies the EQOS Asia equation to a real case, and you can ask which inputs move your score most (EQOS Asia contact page, as of October 2026). The method page describes the 30/60/90 day lever plan that follows an analysis (EQOS Asia method page, as of October 2026).
See also in this guide
Frequently asked questions
What is a sensitivity analysis in simple terms?
It is a test in which one input moves across its plausible range while the others stay fixed. The change in the score shows how much that input matters. The ranking of inputs by swing tells a team where careful sourcing pays off most.
Does EQOS Asia publish the weights used in φ(S)?
The expanded form and the weighting coefficients are proprietary and available under NDA. The nine dimensions, the range from 0 to 1, and the 0.618 threshold are public, so a reader can still reason about the shape of the score (EQOS Asia method page, as of October 2026).
Can I try a sensitivity analysis myself?
Yes. The EQOS Asia simulator offers 9 sliders from 0 to 100, a reset to 50, a random draw, and a target view with the top 3 levers. Moving one slider at a time reproduces the method of this guide on a demo case (EQOS Asia simulator page, as of October 2026).
Why does a low rating move the score more than a high one?
In a geometric product, each rating multiplies the others, so a low rating pulls the whole result down. In the simplified equal-weight example of this guide, a cut of 0.4 lowers the score by 0.082, while a rise of 0.4 lifts it by 0.034.
Test your knowledge
1. How many inputs move at a time in the basic method?
- One
- Three
- All nine
Show the answer
One.
2. In the equal-weight example, what score follows from cutting one rating from 0.5 to 0.1?
- 0.534
- 0.5
- 0.418
Show the answer
0.418.
3. What does the Reset button of the simulator do?
- Sets every slider to 100
- Returns every slider to 50
- Deletes the page
Show the answer
Returns every slider to 50.
4. Which effects does the influence cascade show?
- First-order and second-order effects
- Only price effects
- Only tax effects
Show the answer
First-order and second-order effects.
5. Are the EQOS Asia weights public?
- Yes, in full
- No, they are proprietary and available under NDA
- Only for the Mapping tier
Show the answer
No, they are proprietary and available under NDA.
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 simulator page (eqos-asia.com/simulator.html), method page (eqos-asia.com/method.html), contact page (eqos-asia.com/contact.html), and the WeWork back-test page (eqos-asia.com/case-studies.html), consulted on 21 October 2026. Page updated on 21 October 2026.