Session 3: Manufacture and Process draw, press/calender loading, dryer steam, etc. with time as well as in the cross machine direction (CD non-uniformity). Nevertheless the products that come out of these processes are amazingly "uniform" in terms of meeting basic speci cations, such as basis weight, thickness, moisture content, and other mechanical and physical properties. This is, of course, owing to today's process control systems.A real question of variability and non-uniformity often comes from the end of this manufacturing chain, product performance in end-use. Product performance is a multi-faceted property. It is often not represented by a standard set of product speci cations, but is, instead, expressed by vaguely de ned words, such as runnability, durability, and softness, for example. These performances are normally a system property, and never are a function of a single property, such as strength and surface roughness, as we will see in the subsequent sections. Therefore, in order to enhance performance, i.e., what customers see as values, it is important to understand the properties of product performance (System B in Fig. 1) rst, and then to approach to the process variables (downstream to upstream approach). Otherwise it is extremely dif cult to resolve problems by investigating impacts of one process variables on the performance (upstream to downstream), because a change in one process variable in the upstream normally requires subsequent changes in the multiple process variables in the downstream.Although it is not obvious in the beginning, these performance problems are almost always related to the variability and non-uniformity of paper structures and properties, and thus to pulp and papermaking processes. It is known that a variability spectrum, in length or frequency, of pulp and papermaking process is extremely wide, such as shown in Fig. 2 [8]. The smallest length scale may be those for bres, 1-3 mm, and the longest scale may be 40-50 km for a jumbo reel. However, some of the performance issues are related to even smaller length scales, such as agglomerations of ller pigments which are in micron scales, or to even longer, such as seasonal variations of wood bre stocks which are in a several-month scale. As we see in the following sections, the variability of microscale structures often appears as a stumbling block to resolve some of the very dif cult performance issues. This is a re ection of the fact that our current manufacturing technology still can't control such small-scale structures. (Theoretical developments for describing random bre network structures are reviewed in [9].