Prior research indicates that spatial skills, such as Mental Rotation Skills (MRS), are a strong predictor for mathematics achievement. Other studies have shown that MRS can be improved through training. This paper explores whether a well-known puzzle-oriented tool for building houses with 3D cubes is effective in improving performance in a standardized MRS measure that recorded accuracy and speed. The field experiment took place with 85 year seven (11-12 yr olds) pupils from an independent secondary school in the south of England. We used two conditions in the experiment, with the puzzle-oriented training tool being the intervention condition. The findings show there was a significant effect for accuracy but not for speed. Contrary to prior research our findings did not show any gender effects. The findings and implications are discussed in light of the existing literature around spatial skills.