Make it or buy it? How a distillery decides whether to build its own blending-water plant

Key takeaways

• There are two routes to blending water and they aim at the same spec: build a purification train on site (pretreatment, reverse osmosis, then a polishing step) or buy water finished to specification and delivered. The spec is not negotiable; the supply model is.
• The economics turn on scale and steadiness of demand. Consumable cost per litre of in-house water is genuinely low — running RO ahead of deionisation can “cut costs up to ten times,” with total cost “rarely exceed[ing] 5 cents a gallon” (US Water Systems) — but that figure covers consumables, not capital, labour, QC, effluent consent or a stalled bottling line.
• The costs buyers forget are the unglamorous ones: membrane replacement every few years, resin regeneration with strong acid and caustic (GAPS Water Treatment), the 15–25% of feedwater that leaves as concentrate, and the documentation that proves every batch met spec.

A distillery reducing matured spirit to bottling strength needs water with almost nothing in it, and there are only two ways to get that water: purify it yourself, or buy it already finished. Both routes end at the same target — near-zero minerals, so the water lowers strength and changes nothing else — so the decision is not about quality but about capital, competence and risk. Scale settles most of it: at high, steady volume an in-house reverse-osmosis-plus-polishing train wins on cost per litre; below that threshold, the plant, the chemicals, the testing and the cost of an unplanned stoppage usually outweigh the saving.

What is the choice, exactly?

It helps to be precise about what is being decided. Nobody is choosing between “pure water” and “mineral water” — the purity target is fixed by what dilution has to do. Regulators treat it as a given: US rules for a bottled-in-bond spirit require it be “[r]educed in proof by the addition of only pure water to 50 percent alcohol by volume” (27 CFR § 5.88). Industry describes the goal in the same terms — treatment gives a distiller “blank slate” water to work from (AXEON Water).

So the real question is a supply-chain one:

• Make. Install and operate a treatment train that turns the municipal or well supply into blending water, and own everything that follows — engineering, consumables, monitoring, waste, and the consequences of a failure.
• Buy. Take delivery of water already finished to a documented specification, with a certificate of analysis per batch, and own only storage, handling and goods-in checks.

Every producer answers this eventually, and plenty run a hybrid: in-house treatment as the primary supply, bought-in water as the fallback that keeps the line moving when the plant is down.

What does an in-house blending-water plant actually consist of?

More than a single machine. A working train has four or five stages, and each one is a maintenance item.

Pretreatment removes sediment and, critically, chlorine — activated carbon or a chemical reducing agent, because free chlorine attacks the membrane downstream and chlorinated water carries its own taint risk into a spirit.

Reverse osmosis does the bulk of the work, forcing water through a semi-permeable membrane that rejects most dissolved salts. RO alone typically gets 90–99% of dissolved solids out, which is close but not all the way.

Polishing takes the last fraction, and here there is a second choice inside the first:

• Mixed-bed deionisation (DI) — cation and anion resins that swap dissolved ions for H⁺ and OH⁻. When the resin exhausts it must be regenerated “with strong acid and base solutions to restore its ion exchange capacity,” typically hydrochloric acid at 4–6% for the cation resin and sodium hydroxide for the anion resin (GAPS Water Treatment). That means acid and caustic on site, or a service contract where a vendor swaps exhausted cylinders for fresh ones.
• Electrodeionisation (EDI) — a “continuous, chemical-free process of removing ionized and ionizable species from feedwater using DC power,” capable of “up to 18 MΩ-cm high-purity water.” Its appeal is what it removes from the site rather than the water: EDI “eliminates the need to store and handle the hazardous chemicals used for resin regeneration in mixed beds,” and DuPont positions it as “a smart alternative to — and effective replacement of — conventional mixed bed ion exchange” (DuPont).

Storage and distribution close the loop — a tank, a recirculating ring main, filtration and often UV. This is the stage most often under-specified, because purified water is chemically aggressive and biologically vulnerable; it does not stay at spec on its own.

The sequencing matters commercially, not just technically. Putting RO ahead of DI is what makes the polishing step affordable: it removes most of the load before the resin ever sees it, so “you can easily cut costs up to ten times, if you use reverse osmosis ahead of DI” (US Water Systems).

What does it cost to run — and where are the costs people miss?

The marginal cost of in-house water is genuinely low. With RO ahead of DI, “even when adding in the cost of RO membranes and filters, the total cost per gallon rarely exceeds 5 cents a gallon” (US Water Systems). That is the number that wins internal arguments — and it is also the number that misleads, because it is a consumables figure. The full cost of ownership includes several lines that never appear on a per-gallon calculation:

Two consumable lines deserve naming because they arrive as step changes rather than a smooth monthly cost. Membranes age. Elements more than roughly three to four years old, or heavily fouled, often will not respond to cleaning, and replacement becomes the more economic answer (AMPAC USA). And on-site regeneration is not free time: as one supplier puts it plainly, “if the DI resin is regenerated on site, this can add both time and expense to the process” (US Water Systems). A service-exchange contract converts that into a predictable operating cost, at a higher unit price — which is, in miniature, the same make-or-buy trade being made at the level of the whole plant.

What do you give up by buying water in?

Honesty cuts both ways here. Buying finished water means accepting a higher price per litre, a lead time, a delivery schedule you do not control, warehouse space for stock, and dependence on a third party’s continuity. For a producer running high volume every day of the week, those are real disadvantages and the in-house case is strong.

What you get in exchange is narrower exposure. The capital stays unspent. The purity spec arrives as somebody else’s obligation, evidenced per batch — and a specification you can hold a supplier to is a different risk profile from a specification you must maintain yourself, at 2 a.m., with a fouled membrane. No acid or caustic is stored on your site. No concentrate stream sits on your effluent consent. And the expertise that keeps ultrapure water at spec through storage and distribution — a genuinely specialist discipline — is bought rather than hired.

How does scale decide it?

Because the fixed costs are large and the variable cost is small, the answer follows volume. Rough arithmetic makes the shape of it clear. Reducing spirit from a typical cask strength near 60% ABV to a 43% bottling strength takes roughly a third of the finished volume as added water — so every 1,000 litres bottled needs a few hundred litres of blending water, and a modest bottling run measured in thousands of cases a year needs a supply measured in tens of thousands of litres. (The exact figure is not a simple subtraction, because ethanol and water contract when mixed.)

That produces three broad positions:

• Small and craft producers. Reduction-water demand is low and often lumpy — a bottling run every few weeks. A plant sized for the peak sits idle between runs, and idle purified-water systems are precisely the ones that develop microbiological problems. Buying in usually wins.
• Mid-size producers with steady demand. The genuine decision zone, and where hybrids live: RO on site with service-exchange DI cylinders, or in-house treatment with a bought-in contingency stock.
• High-volume bottlers. Continuous, predictable draw amortises the capital quickly, justifies an EDI polishing stage and full-time water-treatment competence, and makes in-house the default.

Steadiness matters as much as size. A plant that runs continuously is easier to keep in spec than one that stops for a fortnight — which is why demand pattern, not just annual litres, belongs in the calculation.

What about the water you throw away?

An RO plant does not convert all its feedwater into product. Recovery — the share that emerges as permeate — generally sits between about 70% and 85% for conventional brackish and industrial RO (AMTA), with the balance leaving as a concentrated reject stream. Where recovery is lower, that stream is substantial: typical recoveries of 50–75% leave “fairly large quantities of the reject (concentrate) stream that needs to be disposed at considerable cost and with significant adverse environmental impacts” (ScienceDirect).

Making water in-house means owning that stream — a discharge consent, possibly a treatment or disposal cost, and a line in your water-use reporting. It does not disappear when you buy instead; it moves to the supplier, who should be able to account for it. Either way it belongs in the conversation, because the industry is measuring itself on exactly this. The Scotch Whisky Association’s work on responsible water use sets sector expectations on efficiency and stewardship rather than treating water as a free input (SWA).

What changes the answer in India?

The regulatory framing does, and it tilts the buy case. In India, packaged water is a food product governed simultaneously by food-safety and standards law: packaged drinking water is covered by IS 14543, and manufacture and sale are tied to certification under the Bureau of Indian Standards scheme, with the ISI mark and its associated factory and testing requirements (BIS Product Manual, IS 14543).

For a buyer, that is useful leverage rather than red tape. A certified, licensed supplier is already operating an audited plant with defined testing obligations, and hands over documentation with each batch. A distillery treating its own water carries the equivalent assurance burden internally — perfectly doable, but it is a quality-system commitment, not just a capital one.

So which should you choose?

Work through these in order; the first few usually decide it.

1. Annual volume and pattern. High and steady favours make. Low or lumpy favours buy.
2. Do you have engineering capability on site? Someone who can diagnose falling recovery and rising conductivity, not just call a contractor.
3. Are you willing to store acid and caustic? If not, the choice narrows to EDI, service-exchange cylinders, or buying in.
4. Do you have a QC function that can own the water spec? Continuous conductivity monitoring plus periodic laboratory confirmation, with records.
5. Can you discharge the concentrate? Check the consent before the capital request.
6. What does an hour of stopped bottling cost? That number sets how much redundancy — or contingency stock — you need either way.
7. Have you priced the hybrid? In-house primary with a bought-in fallback is often cheaper than in-house with full redundancy.

There is no universally correct answer, and any supplier claiming otherwise is selling. What is universal is the specification: whichever route delivers it, the water must reach the reduction tank near-mineral-free, documented, and biologically clean — because that is the only version of it that changes the strength of a spirit and nothing else.

Glossary

• Blending (reduction / proofing) water — near-mineral-free water used to cut a spirit from cask strength down to bottling strength.
• Reverse osmosis (RO) — purification that forces water through a semi-permeable membrane, rejecting most dissolved salts; the bulk-removal stage of most trains.
• Deionisation (DI) / mixed bed — ion-exchange resin that swaps dissolved ions for H⁺ and OH⁻, which combine into pure water; used to polish RO permeate. Exhausted resin must be regenerated with strong acid and caustic.
• Electrodeionisation (EDI) — continuous, chemical-free polishing that uses DC power instead of regenerant chemicals.
• Recovery — the proportion of feedwater an RO system delivers as product (permeate); the remainder leaves as concentrate.
• Concentrate / reject — the salt-enriched stream an RO plant discharges, requiring a disposal route and consent.
• Service exchange — a supply model in which a vendor delivers fresh DI cylinders and removes exhausted ones for off-site regeneration.
• Certificate of analysis (CoA) — batch document stating measured values against specification; the evidence a bought-in water met spec.
• Total cost of ownership — capital plus consumables, labour, QC, waste and downtime risk, as distinct from a per-litre consumables cost.

Frequently asked questions

Is in-house blending water cheaper than buying it? Per litre of consumables, almost always — with RO ahead of DI, total cost “rarely exceed[s] 5 cents a gallon” (US Water Systems). On total cost of ownership it depends entirely on volume, because capital, labour, QC, waste handling and downtime risk are largely fixed and must be spread across however much water you actually use.

Is bought-in water lower quality than water made on site? No — quality is a function of the specification and the evidence behind it, not of who made it. A documented spec with a batch certificate of analysis is auditable in a way an unmonitored in-house plant is not. The failure mode to watch in both cases is the same: purified water losing spec during storage and distribution.

Do I need EDI, or will mixed-bed DI do? Both can polish RO permeate to blending-water purity; EDI’s advantage is operational rather than chemical, because it removes regenerant acid and caustic from the site and runs continuously on electricity (DuPont). Mixed beds remain a sound choice, particularly on a service-exchange contract.

How much water does an RO plant waste? Typically 15–30% of feedwater leaves as concentrate, with conventional brackish and industrial systems recovering roughly 70–85% (AMTA). At lower recoveries the reject volume becomes significant enough that disposal carries real cost and environmental impact (ScienceDirect).

Can I do both? Yes, and many producers do. Running an in-house train as primary supply with a stock of bought-in blending water as contingency is often cheaper than engineering full redundancy into the plant — and it means a fouled membrane delays a maintenance job rather than a bottling run.

Blenders International · Water Science series