Catalytic reduction with Pd has emerged as a promising technology to remove a suite of contaminants from drinking water, such as oxyanions, disinfection byproducts, and halogenated pollutants, but low activity is a major challenge for application. To address this challenge, we synthesized a set of shape- and size-controlled Pd nanoparticles and evaluated the activity of three probe contaminants (i.e., nitrite, N-nitrosodimethylamine (NDMA), and diatrizoate) as a function of facet type (e.g., (100), (110), (111)), ratios of low- to high-coordination sites, and ratios of surface sites to total Pd (i.e., dispersion). Reduction results for an initial contaminant concentration of 100 μM show that initial turnover frequency (TOF0) for nitrite increases 4.7-fold with increasing percent of (100) surface Pd sites (from 0% to 95.3%), whereas the TOF0 for NDMA and for diatrizoate increases 4.5- and 3.6-fold, respectively, with an increasing percent of terrace surface Pd sites (from 79.8% to 95.3%). Results for an initial nitrite concentration of 2 mM show that TOF0 is the same for all shape- and size-controlled Pd nanoparticles. Trends for TOF0 were supported by results showing that all catalysts but one were stable in shape and size up to 12 days; for the exception, iodide liberation in diatrizoate reduction appeared to be responsible for a shape change of 4 nm octahedral Pd nanoparticles. Density functional theory (DFT) simulations for the free energy change of hydrogen (H2), nitrite, and nitric oxide (NO) adsorption and a two-site model based on the Langmuir–Hinshelwood mechanism suggest that competition of adsorbates for different Pd sites can explain the TOF0 results. Our study shows for the first time that catalytic reduction activity for waterborne contaminant removal varies with the Pd shape and size, and it suggests that Pd catalysts can be tailored for optimal performance to treat a variety of contaminants for drinking water.