Do the arithmetic. Åreskutan rises a little over a thousand metres above lake Åresjön. A twelve kilometre window compressed into 25 centimetres gives a scale of roughly 1:50,000. A thousand metres then becomes two centimetres.
Two centimetres is not nothing. But it is not a mountain either. It is a bump on a tray, and once you hang it on a living room wall and look at it head on under ordinary ceiling light, it nearly vanishes.
That is why essentially every relief map exaggerates height. The old plaster models at school did it, relief globes do it, and so do we.
Why the eye needs the exaggeration
We do not read terrain by measuring height. We read it through shadow.
A slope is visible because it is darker than the surface next to it. How dark it becomes depends on the gradient, not on the height. At 1:1 scale most real slopes end up so gentle in the model that they barely cast any shadow at all, and then all the information at middle distance disappears. Middle distance is exactly where a mountain becomes recognisable: not at the summit point, but in the ridges and valleys in between.
Scale the height up and the gradients get steeper, the shadows deeper, and the shape readable right across a room. That is the whole point of making a map physical.
What the exaggeration must not touch
And here is the important boundary. The exaggeration is vertical.
The plan is not touched. Every valley lies where it lies, every bend in a ridge has its true shape seen from above, the distance between two summits is correct. Only one axis is scaled, identically across the whole surface. A steep slope and a gentle slope keep their relationship to one another.
That is the difference between exaggeration and invention. No contour is moved, no summit is added, nothing is smoothed out to make it prettier.
How much is right?
It depends entirely on the place, which is why it is a craft decision rather than a setting.
A steep massif needs almost nothing. Mont Blanc has so much elevation range over such a short distance that it barely needs help at all; the relief stands on its own. A rolling Swedish mountain landscape needs more, otherwise the whole square becomes one evenly tilted plane.
There is also a limit upwards. Push the height too far and the slopes go nearly vertical, small irregularities in the data turn into spikes, and the mountain stops looking like itself. There is a practical reason too: every piece in the same range should read at roughly the same scale. Hang a Trysil piece next to a Mont Blanc piece and they should look like they come from the same world. So we keep the whole range within a fairly narrow band rather than maximising each subject on its own.
The test we actually use
The decisive test is not a number. It is walking five metres away and seeing whether the place is recognisable.
A relief that is geographically perfect but reads as a shapeless lump has failed at what it is meant to do. That is a harder requirement than it sounds, and it is also why we do not make every place we are asked for. Some landscapes are simply too flat to carry a relief, however beautiful they are in person.
More on what decides that in the guide to elevation models. And to see the exaggeration in practice, the edge view of the Åre piece is clearest: it shows how far the terrain actually stands out from the backing plate.