Membrane Procurement: The Lock-In Written Into the Tender

A membrane treatment plant is bought once and re-bought continuously. Elements are replaced on a cycle that typically runs three to seven years depending on feed quality, pretreatment performance and operating regime, and membrane replacement commonly accounts for something in the order of five to ten percent of annual operating cost. Over a twenty-five year asset life that is three or four full replacement events, each one a procurement exercise in its own right. Almost none of that spend is competitively tendered, because the moment at which the buyer holds maximum leverage is the moment at which membranes are not on the table.

At capital tender the buyer controls award, has multiple bidders in play and can attach conditions to anything. Replacement elements sit outside the capital scope, so they are not priced. By the time the first replacement is due, the plant is operating, the performance guarantee is in force, the operator is familiar with one product, and the buyer is negotiating with a single supplier against an asset that cannot stop. The leverage inverted while nobody was watching, and it inverted at award.

Standardisation that does not mean interchangeability

The eight-inch spiral-wound element is a nominally standard form factor. It fits standard pressure vessels from multiple manufacturers, and that geometric compatibility persuades a great many buyers that they have a competitive market available to them whenever they want it. Geometry is the least important part of the question.

What actually constrains substitution is the projection model. Membrane plants are sized using the membrane supplier's own projection software, and the flux distribution, stage recovery, array configuration, feed pressure and permeate quality in the design case all come out of that model. The performance guarantee is written against those outputs. The acceptance test is written against the guarantee. Fitting a competitor's element into the same vessel is trivially possible and contractually awkward, because the element that arrives has a different permeability and a different rejection characteristic, which changes the pressure profile across the array, which puts the guarantee into dispute. The supplier who wrote the model is under no obligation to stand behind a stage populated with somebody else's membranes, and no operator wants to be the person who voided the guarantee to save a percentage on consumables.

In membrane bioreactor plants the position is considerably worse and is frequently misread as being the same. MBR cassettes, modules, racks and their aeration arrangements are proprietary. There is no common form factor, the tank geometry is designed around one supplier's module dimensions, and the permeate and air manifolds are built to match. A municipal authority that awards an MBR plant has not selected a technology. It has selected a supplier for the design life of the concrete.

Where the cost is transferred, not removed

Membrane life is largely a function of what reaches the membrane. Pretreatment determines fouling rate, fouling rate determines cleaning frequency, cleaning frequency determines irreversible flux loss, and flux loss determines when elements come out. This chain is well understood by every process engineer and is routinely defeated by tender economics.

Under-specifying pretreatment is the single cheapest way to win a competitive water treatment tender. It reduces capital cost visibly, in the column the evaluation panel is scoring, and it moves the consequence into a recurring consumable line that appears years later, in a different budget, under a different manager, attributed to feed water variability rather than to a procurement decision. The bid that wins is frequently the bid that transferred the most cost across that boundary, and the evaluation method rewards it because the evaluation method compares capital.

The specific failure is that the interface between pretreatment and the membrane racks is usually left undefined. If the contract does not state the water quality that must be delivered to the membrane feed, expressed in terms the parties can measure, then when membrane life falls short of expectation there is no way to establish whether the pretreatment underperformed or the membranes did. Both suppliers point at the feed water. The owner pays. This is not a dispute that can be won after the fact, because the evidence that would have settled it was never specified into existence.

How we structure a membrane package

At D1R7K0N we treat the replacement cycle as part of the original procurement rather than as a future operating matter, because that is the only point at which it can be priced competitively. Four things go into the enquiry before award.

Priced replacement cycles. Bidders quote the first and second full replacement sets with a defined escalation mechanism, submitted with the capital bid and evaluated alongside it. This does not commit the owner to buying from that supplier. It establishes a ceiling and a reference price at the only moment when competitive tension exists, and it exposes the bidder whose capital number is low because the consumable number is high.

The projection model as a deliverable. The design projection, its input assumptions, the fouling and flux decline allowances used, and the software version are contract deliverables held by the owner. Without them the owner cannot independently evaluate an alternative element, cannot verify whether an underperformance claim is genuine, and cannot brief a second supplier at replacement time. With them, the owner can put a replacement set out to competitive enquiry and have the responses assessed against a common basis.

A defined feed quality boundary. The specification states the water quality that pretreatment must deliver to the membrane feed, using measurable fouling indicators, and locates the responsibility for membrane life on the correct side of that boundary. Where a single contractor supplies both pretreatment and membranes, this matters less commercially and still matters operationally. Where they are separate packages, it is the difference between a warranty claim and an argument.

Interchangeability stated as a requirement. For pressure-driven RO and nanofiltration, we require standard-envelope elements and ask the bidder to name qualified alternates that the guarantee will tolerate. Where a design depends on a proprietary element or a proprietary module, that dependency is identified in the technical evaluation as a commercial risk with a number attached, not left as a footnote in a datasheet. Sometimes the proprietary product is the right answer on process grounds. It should be selected knowing what it costs over twenty-five years, not by default.

The question to ask before award

The useful test on any membrane plant, whether it is a seawater desalination train, a municipal MBR or an industrial reuse system, is a simple one. If the selected supplier doubled its element price at the first replacement, what would the owner do? If the honest answer is pay it, the procurement has already concluded and the plant has not yet been built.

That answer can be changed, but only before award. Afterwards the tank is poured, the vessels are installed, the guarantee is signed and the operator has been trained on one product. Membrane technology selection is a process engineering decision and it is treated as one. The replacement market that selection creates is a procurement decision, and it is usually not treated as anything at all.