Non-linear value functions and diminishing returns in scoring criteria
How non-linear value functions and diminishing returns work in scoring criteria: shape a value curve, normalize scores, and read a multi-criteria total fairly.
Contents
- What a value function does in scoring
- Linear versus non-linear scoring explained
- Diminishing returns in scoring criteria
- Concave and convex value curves side by side
- How the curve shape changes a shortlist
- Building a value curve for each criterion
- Weights, totals and the multi-criteria value function
- Mistakes to avoid with non-linear scoring
- Questions about value curves and diminishing returns
- What does diminishing returns mean in scoring criteria?
- How does a value function work in multi-criteria scoring?
- What is the difference between linear and non-linear scoring?
- When should a criterion use a convex curve?
- How does score normalization relate to the value curve?
- Why does the shape of a value curve change a shortlist?
A non-linear value function is a scoring curve that converts a raw measurement into a value score by unequal steps along the scale. Diminishing returns in scoring criteria means each extra unit of a measure adds less value than the unit before, so the curve bends. Getting that shape right keeps a multi-criteria score honest about what extra performance is actually worth.
- A linear score treats every extra unit as equally valuable, which flatters criteria with headroom.
- A concave curve captures diminishing marginal value: early gains count more than later gains.
- A convex curve models rising value, where high performance matters far more than small gains.
- Score normalization strips units from measurements before the value curve sets the spacing of scores.
- Curve shape decides winners, so set it before you compare listings, not after.
What a value function does in scoring
A value function maps a measured result onto a score that reflects how much that result is worth to the decision. The map can be a straight line or a curve, and the difference changes the ranking of options. A value function sits under each criterion and feeds the multi-criteria total.
The utility function value curve borrows from decision theory, where a utility function curve describes how a person trades off outcomes against each other. In a scoring sheet the same idea becomes practical: you state what a jump from one score to the next is worth. That statement is a judgment, not a measurement, and it belongs in writing.
Criteria rarely share a unit, so each measurement passes through score normalization before the curve applies. Normalizing to a fixed range lets you apply one curve per criterion without mixing meters, hours and money. The value function shape then controls how scores are spaced inside that range.
Linear versus non-linear scoring explained
Linear scoring gives every extra unit of a measure the same value from the bottom of the scale to the top. Linear versus non-linear scoring is the first choice in any scorecard, because a straight line asserts equal worth at every level. That assertion is rarely true across a full range.
A non-linear scoring curve keeps the direction of the criterion but changes the spacing of scores. The curve can reward early improvement or reserve high marks for exceptional results. Either way, the reader of the shortlist sees a score that reflects a stated judgment rather than a raw number.
A linear score also makes criteria with large ranges dominate a total without anyone deciding that they should. Non-linear curves let you compress a range where gains stop mattering. The scoring criteria then carry the weight you intend.
Diminishing returns in scoring criteria
Diminishing returns in scoring criteria appear when each added unit of a measure raises its value score by a smaller step. The curve bends downward as the measure rises, and this shape is called concave. A concave curve protects a total from a single runaway figure.
Diminishing marginal value has a plain reading: the first hour of a service saves the buyer more trouble than the tenth. A concave curve encodes that reading in the scoring rule. An option that is already strong gains little from a further improvement, which is the point.
Concave shapes suit criteria with a practical ceiling, such as a rated capacity or a supported standard. Beyond the point where the need is met, more of the measure adds little to the decision. Set that point from your own requirements and state it next to the curve.
Concave and convex value curves side by side
Concave and convex value curves bend in opposite directions and reward opposite ends of a range. A concave curve favors modest, steady performance across several criteria. A convex curve favors an option that reaches high performance on one criterion while giving up little elsewhere.
A convex curve models rising value, where each added unit is worth more than the last. That shape fits a criterion with a threshold, such as a safety margin or a compatibility level. Below the threshold the score stays low, and above it the score climbs quickly.
Risk-averse utility explains why a buyer might choose the concave curve even when the average outcome is similar. Diminishing marginal value makes an extreme result less attractive than a reliable middle result. The curve shape is where that preference enters the scorecard.
How the curve shape changes a shortlist
The value function shape decides which listing tops a shortlist, because it sets how much each strength counts in the total. Two criteria can carry equal weight and still trade places under different curves. Curve shape is a ranking decision, not a formatting detail.
The marginal value of criteria shifts along the scale, so a curve can make one criterion important at low scores and minor at high scores. A weighted sum with straight lines cannot show that shift. A curve makes the shift visible and repeatable across options.
Keep every listing side by side and apply the same normalized curve to each. Comparing listings under one fixed rule lets you see which option wins on shape alone. Changing the curve after you have a favorite turns the exercise into a search for a justification.
Building a value curve for each criterion
A single-criterion value curve is built one criterion at a time, starting from the range you expect to see. Fix the minimum and maximum, then decide what a middle result is worth. Three anchor points are enough to sketch a curve, and the anchors keep the judgment explicit.
Choose the shape from the meaning of the criterion rather than from the data in front of you. A criterion with a practical ceiling takes a concave curve, and a criterion with a threshold takes a convex curve. Record the reason in the scorecard so another reader can check the logic.
After the curves are set, apply score normalization so the raw units do not leak into the total. Fit each listing's measurements to the same anchors and read the value score. A total that rests on stated anchors can be explained in one sentence to anyone reading the shortlist.
- Fix the minimum and maximum you expect for the criterion.
- State the value of a middle result as a third anchor.
- Pick concave for a practical ceiling and convex for a threshold.
- Normalize the raw measure before applying the curve.
- Write the reason for the shape beside the criterion.
- Apply one curve to every listing before comparing totals.
Weights, totals and the multi-criteria value function
A multi-criteria value function combines single-criterion scores into one total through weights you set. Weights say how much each criterion matters, while curves say how much each level of performance matters. Confusing the two roles produces totals that cannot be explained.
A criterion can carry a low weight and still swing the total through a steep curve. Check each criterion's curve and weight together to see the true influence. If a criterion dominates, say so and lower its weight or flatten its curve.
Keep the number of criteria small enough that each curve can be defended. A long list of criteria with sharp curves turns the total into noise. A short list with clear shapes lets the reader follow the reasoning from measurement to score.
Mistakes to avoid with non-linear scoring
A curve fitted to past choices rewards what you already preferred and hides the judgment. Set anchors from requirements rather than from the listings you happen to hold. A stated reason for each shape keeps the scorecard open to challenge.
Mixing normalized and raw scores in one total breaks the spacing the curve provides. Apply the curve after normalization and before weighting for every criterion. A single raw number inside a curved total distorts the result without warning.
Steep convex curves on safety or suitability criteria can push unsafe options up the list. Keep safety as a pass or fail gate rather than a scored criterion. Where a judgment needs a trained specialist, ask for that assessment instead of encoding it in a curve.
Questions about value curves and diminishing returns
What does diminishing returns mean in scoring criteria?
Diminishing returns in scoring criteria means each added unit of a measure adds less value to its score than the unit before. The curve bends downward as the measure rises, and this concave shape keeps a strong result from swamping the total. The curve states that early gains matter more than later gains.
How does a value function work in multi-criteria scoring?
A value function in multi-criteria scoring converts each measurement into a score that reflects how much that result is worth. Each criterion gets its own curve, and weights then combine the scores into one total. The curve shape and the weight play separate roles in that total.
What is the difference between linear and non-linear scoring?
Linear scoring gives every extra unit of a measure the same value across the range, while non-linear scoring changes the spacing of scores. A curve can reward early improvement with a concave shape or reserve high marks with a convex one. Non-linear scoring lets a scorecard reflect a stated judgment about worth.
When should a criterion use a convex curve?
A criterion uses a convex curve when each added unit is worth more than the last, which fits a threshold such as a compatibility level or a safety margin. Scores stay low below the threshold and climb quickly above it. Below the threshold the option earns little credit for small gains.
How does score normalization relate to the value curve?
Score normalization converts each measure to a fixed range before the value curve applies, so units do not distort the total. The curve then sets the spacing of scores inside that range. Normalizing first lets one curve shape work with meters, hours or money.
Why does the shape of a value curve change a shortlist?
The shape of a value curve changes a shortlist because it sets how much each strength counts when criteria are weighted. Two listings can trade places under a concave curve and a convex one with the same weights. Fixing the curve before comparing listings keeps the ranking honest.