The Kepler Mission was designed to measure the frequency of Earth-like planets in the habitable zone of Sun-like stars. A requirement for determining the underlying planet population from a sample of detected planets is understanding the completeness of that sample—what fraction of the planets that could have been discovered in a given data set were actually detected. Here we describe an experiment designed to address a specific aspect of that question, which is the issue of signal throughput efficiency. We investigate the extent to which the Kepler pipeline preserves transit signals by injecting simulated transit signals into the pixel-level data, processing the modified pixels through the pipeline, and measuring their detection statistics. For the single channel that we examine initially, we inject simulated transit signal trains into the pixel time series of each of the 1801 targets for the 89 days that constitute Quarter 3. For the 1680 that behave as expected in the pipeline, on average we find the strength of the injected signal is recovered at 99.6% of the strength of the original signal. Finally we outline the further work required to characterise the completeness of the Kepler pipeline.