I am, though, doubtful about the value of a logarithmic scale in this case. I do not think that a bar graph should use this sort of a scale unless it is intended as a really rough visualisation technique.

I guess that a simple table with a list of categories in acs or desc order would server a similar purpose, use less real-estate and possibly be less confusing…

Would be interesting to hear other people's thoughts about the value of logarithmic scales, so please comment :)

Consider an axis mapping function Y(y) = log^10(100 * y) / 2. This will effectively lift the scale of the logarithm above your fractional values that are less than 1, then provide a linear correction to compensate for the "scale up".

Yes, I considered it, tried it and failed. Unfortunately, Logarithmic scale works quite bad for values under 1 (say 0.1) – they show as a negative. Especially annoying when you have to present percentages and some of them are really small.

Have you considered presenting the data from the pie chart on a logarithmic scale rather than a scale with discontinuities in the axis? Most moderately mathematically literate users can read a log chart effectively, and it gives you a clearer view of relative scale than a discontinuous axis. The only downside is that log charts really need gridlines in the background.

Sadly your story rings a bell… :(

Also, the second bar graph is somewhat pointless – since the users of the report cannot see how the expenditure amouints are distributed per focus. They can just see how focus expenditure is distributed per category..But everything looks even – therefore, again, they look good, because the allocations look quite even, while in reality there are massive gaps..

Cheers,
Nick