Classically economics does not admit goods that have increasing marginal utility. There are a few arguable exceptions (drug addiction is a big one), but mostly it's held that they don't exist.
But consider an office thermostat. From -20 degrees Celsius to -10, there is little gain in marginal utility. No one will work there either way. From 10 degrees Celsius to 20, there is a massive utility increase as the office goes from too chilly to comfortable.
Marginal utility obviously increases... So what gives?
Utility, in the Mengerian tradition, does not refer to sensation or satisfaction but to how useful something is to achieve a given end. So marginal utility means how an additional unit of a homogeneous good ranks in terms of its usefulness in achieving that end. So if you are hungry and you want to remove that hunger, the first bite of a sandwich is ranked as most useful because you need to have it in order to begin removing your hunger, even if the second bite is more physically or psychologically satisfying than the first. That first bite is ranked higher than the second because it gets you closer to your goal, and so on.
Therefore, the Mengerian solution to our problem is to ask :
- where am I now (-20C)
- where do I ideally want to be (+20C).
When you choose to increase the temp by Δ10C (from -20C to -10C), while -10C may not feel that much different to you from -20C, you would still rank that Δ10C higher than the next Δ10C because you need to get from -20C to -10C before you can, by adding another Δ10C, get from -10C to 0C, even though that second Δ10C difference feels significantly better to you than the first Δ10C; and so on until you reach your subjectively desired goal of +20C, where you feel the most comfortable. If you then go another Δ10C to 30C, the marginal utility continues to fall because it takes you farther, not closer, to your goal.