Your character’s defensive stat rises every time you buy an upgrade, yet the next purchase seems less noticeable than the first. That can happen when the game’s formula gives each additional point a smaller effect. The total is increasing while the marginal benefit is shrinking.
The following numbers are an invented teaching example, not a formula claimed for any particular game. Suppose a fictional damage-reduction system uses stat divided by stat plus 100. A stat of 100 gives 50% reduction; 200 gives about 66.7%; 300 gives 75%.
Compare the next step, not only the final total
In that example, the move from 100 to 200 adds about 16.7 percentage points of reduction. The move from 200 to 300 adds about 8.3 points. Equal increases in the displayed stat do not produce equal increases in the displayed percentage.
That does not automatically make the later upgrade bad. Its value depends on what the resulting reduction means in play, the upgrade’s cost and the alternatives. Percentage points, relative damage reduction and effective durability are different ways of describing an outcome, so identify which one you are comparing.
Real games may use caps, breakpoints, additive bonuses, multiplicative bonuses or mode-specific rules. Read the current game’s documentation or test carefully before applying a formula found for another title or an older patch.
Make a small upgrade comparison
Record the current build, proposed change, resource cost and measurable outcome. Keep other equipment and test conditions stable where possible. Compare the benefit of the next purchase with another use of the same resources, rather than ranking upgrades solely by the largest stat number.
For a group build, also consider whether the change improves consistency or solves a specific encounter problem. A modest numerical gain can matter when it crosses a meaningful threshold, while an impressive total can be irrelevant to the task.
Diminishing returns is a question about the relationship between input and benefit. It is not a universal instruction to stop investing after an arbitrary number of upgrades.