A central goal in microbial ecology is to simplify the extraordinary biodiversity that inhabits natural environments into ecologically coherent units. We profiled (16S rRNA sequencing) > 700 semi-aquatic bacterial communities while measuring their functional capacity when grown in laboratory conditions. This approach allowed us to investigate the relationship between composition and function excluding confounding environmental factors. Simulated data allowed us to reject the hypothesis that stochastic processes were responsible for community assembly, suggesting that niche effects prevailed. Consistent with this idea we identified six distinct community classes that contained samples collected from distant locations. Structural equation models showed there was a functional signature associated with each community class. We obtained a more mechanistic understanding of the classes using metagenomic predictions (PiCRUST). This approach allowed us to show that the classes contained distinct genetic repertoires reflecting community-level ecological strategies. The ecological strategies resemble the classical distinction between r- and K-strategists, suggesting that bacterial community assembly may be explained by simple ecological mechanisms.