 We have developed two new methods for improving laser-based forest inventories. Firstly, by using last-pulse data, instead of the first-pulse data, we are able to detect trees more accurately. This is especially useful when there are many trees overlapping one another as the last-pulse data provides a clearer picture of the terrain. Secondly, we have combined the traditional point-height metrics with individual tree-based features to improve the accuracy of predicting forest variables such as volume and diameter. Using these two methods together has resulted in significant improvements in the accuracy of our predictions. This article was authored by Hanu Haipa, Miko Vasturanta, Marcus Holopaynon, and others.