Quad analysis is useful for analyzing subjective properties. This new approach provides an efficient route to conducting paired comparisons, the preferred method for quantifying such properties. A process study taken from the carpet manufacturing industry demonstrates the usefulness of the approach: drawtexturing factors are manipulated to impact commercially important carpet properties like body, tuft endpoint. and wearability. Quad analysis provides process direction, whereas traditional sample-to-control contrasts fail to provide any useful process direction. An extension to quad analysis. quad folding of one quad design into another, is introduced as an efficient method to rank order the subjective properties of larger data sets. Its advantage is demonstrated for a series of nylon carpets inserted with a low-melt yam to improve their initial presentation and wearability.No matter how it might wish to rely on objective testing, the carpet industry-and for that matter, the textile industry in general-often relies on the subjective opinions of its expert graders to quantify carpet properties, e.,y., hand, body, harshness, texture quality, endpoint definition, and wearability. Typically, graders rank one or more of these properties: a sample carpet against a control carpet or against a series of controls. While these analyses assess sample-to-control differences adequately, all too often they offer little indication of just how much the samples differ among themselves, and as a result, give little process direction. The object of such an exercise is many times imbedded in the sample-tosample (not sample-to-control) relationships. so the former's imprecise measure is problematic.Subjective comparisons by human graders also suffer error: from a grader's competence, from sample bias, and from the nonlinearity of a grader's sensory responses. Limitations associated with the varied approaches to subjective testing have been discussed in the literature [ 13]. In the absence of objectively defined measures, the recommended approach has been either a simple rank ordering or paired comparisons. Rank ordering a large group of samples (more than five). however, presents problems as well, the most important of which is that it's s not effective unless the differences between items are large [3]. With their obvious downside, either head-tohead contrasts against a control or forced rankings are conducted instead of paired comparisons for one reason: speed of assessment. The paired-comparison approach with its superior discrimination would be used more frequently if it were faster. but it's just too time consuming for most commercial manufacturing operations.As well, more precise measures of subjective properties are needed to better extract process direction from factorial experimentation. Contrasting items against a control, though sometimes valuable. doesn't differentiate well enough to extract the relationships between test -items. On the other hand. rank ordering a large group is imprecise and provides no assessment of sta...