The performance of the Kenics static mixer for mixing small streams of passive tracer into the bulk flow is investigated as a function of injection location and flow ratio. Flow ratios of 1/99 and 10/90 are simulated at nine different injection locations, and two alternative geometries are considered in addition to the standard Kenics mixer. Mixing is evaluated qualitatively by examining the spread of the tracer on cross-sectional slices from the mixer and quantitatively by computing the variation coefficient as a function of axial position. For the standard Kenics geometry, injection location strongly affects the extent of mixing only for the first few elements, after which the mixing rate is independent of injection location. In a sufficiently long mixer, material injected at any location spreads to the entire flow, but the least effective injection locations require up to four elements more than the most effective locations to achieve the same variation coefficient. A faster rate of decrease in variation coefficient is observed for a flow ratio of 1/99 us. 10/90. An alternative geometry in which the elements have 120° of twist instead of the standard 180° of twist shows a similar dependence on injection location and flow ratio, but is more energy-efficient than the standard Kenics geometry. In another alternative geometry in which all elements have the same direction of twist, segregated islands exist in the flow. For injection locations inside the segregated islands, virtually no mixing takes place; for injection locations outside of the segregated islands, the tracer spreads to the remaining flow but does not penetrate the islands.