In principle, it's easy to solve a data-acquisition and analysis problem using a single-path signal chain which goes from sensor input, to A/D converter, to processor. In some cases, though, a parallel analog-path approach, along with some analog pre-processing, makes more sense in terms of cost, power, and reliability.
Think about a system designed to monitor a bearing assembly. By continually analyzing the vibration-sensor output, the system can determine the bearing's condition and any degradation due to wear, lubrication, heat, loading, and other problems. At the same time, the monitor might be part of a sudden-shutdown alarm function which shuts off the power if the load stalls, such as when the machine jams.
The single-path approach to sensing and monitoring requires a fairly fast, high-resolution converter and processor to digitize and analyze the data stream. Instead, a better approach may be to look at this as two problems: one of slower trend assessment, and another of fast alarming. The former needs high accuracy and resolution, the latter needs only a modestly accurate threshold.
For these reasons, it may make more sense to look at another design approach. Use a basic analog comparator with settable threshold for near-instantaneous monitoring of sudden overloads and stalls. It's also very reliable, since there is relatively little circuitry or software involved. For the digital path, consider pre-processing with an rms-to-dc converter ahead of the A/D converter and processor, to produce a continuous signal corresponding to the sensor output's average value. The dual combination lets you to use slower, lower-power, smaller devices, compared to a simplistic fast digitize/analyze approach.
In other words, a hybrid architecture lets you match signal paths and implementations to the data needs. You get a design which is also more reliable, even though the BOM has more parts, which is a bad thing at first glance. But first looks can be deceiving, so it pays to step back and ask what your needs and priorities really are.
-x x x-