Hi Salvi,
From my experience of using Airwave VisualRF so far is:
1) For office environment (in real life), the predictive heatmaps are pretty similar in shapes. But, there is around 10dB difference between what we see in a real site survey compared to the predictive heatmap in Airwave. The predictive heatmap in Airwave is set at default environment (office).
2) Draw walls to the best knowledge will give more accurate predictive results
3) Places where we see coverage less than the light blue (lower than -75db) are place of concern. This gives a good idea on how big is the area that we are trying to cover and where we are lacking in coverage without having to send someone to do a physical site survey.
4) We have done around 3-4 physical site survey in different locations and compare it with the predictive heatmaps and they are pretty close (i would say consistently with about 10dB difference).
Besides this, we are compiling our own metric to determine Access Point which might indicate problems with coverage using the client sessions data by looking at the client signal strength.