On-line monitoring of fluctuations in pressure measurements recorded near the die offer significant potential for tracking paste quality in continuous paste extrusion systems. Several signal processing techniques, developed previously for ram extrusion applications, have been applied to a continuous, twin-screw paste extruder processing a clay-based detergent paste. These techniques include the removal of corrupting equipment noise; outlier analysis to detect pockets of entrapped air and large agglomerates in the paste; and quantification of the signal noise, either by fractal analysis or standard error analysis. The signal noise is related to the overall homogeneity of the paste, which can be influenced by factors including dry powder mixing, particle size distribution and paste hold-up within the barrel. The signal processing results are compared with physical measures of extrudate quality, and show a good correlation. Fractal analysis proved a less suitable quantitative tool than the other approaches for this application. The methods are compared with traditional statistical process control techniques and future application is discussed.