The best known waterless process replaces water with carbon dioxide above its critical point of 31.1 °C and 7.38 MPa. In that state carbon dioxide dissolves non-polar disperse dyes and swells polyester, so that dyeing proceeds without water, salt or drying, and the gas is recycled (Saus et al., 1993; Bach et al., 2002). The process is used industrially for polyester. Cotton is another matter. Cellulose is polar and is not swollen by carbon dioxide, and disperse dyes have no affinity for it. Research has pursued reactive disperse dyes that can bond to cellulose from the supercritical medium, chemical modification of the fibre, and polar co-solvents, with encouraging laboratory results but without an established industrial process (Fernandez Cid et al., 2005; van der Kraan et al., 2007; Banchero, 2013). High-pressure vessels also carry a capital cost of a different order from that of an atmospheric dyeing machine (Bach et al., 2002).

For cotton, the routes that reduce water substantially today are low-water and not strictly waterless. Cold pad-batch dyeing applies concentrated dye liquor by padding and fixes it at ambient temperature without salt. Foam and spray application replace most of the water in the liquor with air and lower the wet pick-up of the fabric. Cationic pretreatment of cotton allows reactive dyeing without electrolyte and with less wash-off (Khatri et al., 2015; Hasanbeigi and Price, 2015). The European reference document on best available techniques lists low liquor ratio machinery and optimised rinsing among the measures already proven at scale (European Commission, 2003). Table 1 compares the main routes.

Firms in a cluster gain from proximity through shared suppliers, labour and knowledge (Porter, 1998), and research on clusters in developing economies adds that deliberate joint action, which Schmitz and Nadvi (1999) call collective efficiency, lets small firms do what none could do alone. A common effluent treatment plant is an example. Clean technology raises the same logic. Studies of manufacturing plants in developing countries find that regulatory pressure drives adoption, and that lack of capital, information and technical capacity holds it back (Luken and Van Rompaey, 2008). Each barrier is easier to address collectively.

Diffusion research supplies the remaining concepts. Rogers (2003) identifies five attributes that govern how quickly an innovation spreads: relative advantage, compatibility with existing practice, complexity, trialability and observability. A pressure dyeing plant scores badly on the last four for a small cotton dyer. A shared pilot can supply the trial and the observation that individual firms cannot arrange for themselves.

The cluster first needs to know what it uses. For a representative set of member dyehouses, the working group records water intake, salt, alkali, dye and auxiliary consumption, steam and electricity, and effluent volume and quality, each expressed per kilogram of fabric dyed and broken down by process step and by shade depth. The reference document on best available techniques gives a structure for such an inventory (European Commission, 2003). The baseline shows where water actually goes. In reactive dyeing a large share is used in rinsing and soaping after dyeing, a fact that shapes the choice of remedy.

Each candidate route is then assessed against criteria agreed in advance: the reduction in water, salt and effluent load it offers relative to the baseline; its technical readiness for cotton in the fabric forms the cluster processes; capital and operating cost; the skills and maintenance it demands; and its effect on shade range, fastness and reproducibility as required by buyers. Because the criteria are of different kinds and members will weigh them differently, the screening is suited to a simple multi-criteria method in which weights are set openly and the sensitivity of the ranking to those weights is examined (Belton and Stewart, 2002). Figure 1 shows the intended output: each option placed by its benefit and its readiness, and assigned to one of four courses of action.

The quadrant labels in Figure 1 are illustrative placements based on the published state of each technology, and a cluster would revise them from its own baseline. The essential point of the matrix is that options with large benefits and low readiness are not rejected. They are assigned to joint piloting, where the risk is shared.

Options in the pilot quadrant are tested on a single installation owned or leased by the association and open to all members, who bring their own fabrics and shades. The arrangement follows the reasoning of strategic niche management, which holds that an unproven technology needs a protected space in which users can learn about it before it meets full commercial competition (Kemp et al., 1998). The pilot protocol is fixed beforehand: which fabrics and shades, how many repeats, which colour fastness and levelness tests, and which resource flows are metered. Members must be able to see the results, including the failures.

Pilot data are evaluated on three counts. Technical performance is judged against the buyers' specifications that members already work to. Environmental performance is assessed by life cycle assessment in accordance with the international standards (ISO, 2006a; ISO, 2006b), with the system boundary drawn widely enough to include the electricity for compression or the reagents for fibre modification, so that a saving in water is not bought with an unexamined burden elsewhere. Economic performance is assessed over the life of the equipment and includes the avoided cost of effluent treatment, which in a cluster under strict discharge rules may be the largest single item (Holkar et al., 2016).

Adoption then proceeds in the order the screening suggests. Measures that are ready and cheap are taken up by individual firms at once. Measures that are ready but require new machinery, such as pad-batch ranges, may be better provided as a shared service, with one facility dyeing on commission for several members. High-pressure or otherwise novel plant is added only if the pilot supports it, and then preferably as a common facility. Financing follows the same division: firm-level loans for the first group, and collective borrowing or public co-funding, justified by the reduced load on the common treatment plant, for the others.