Identify the real problem first
Stated goals are rarely the true constraint. You might ask for “potable quality” or “95% TDS removal” but the underlying issues are feed swings, foulants, concentrate disposal, or regulatory limits on specific micropollutants. Start by mapping failure modes: flux loss from organic or biofouling, scaling from hardness and silica, and chemical attack from oxidants. If your requirement includes removal of divalent hardness or specific organics, consider a nanofiltration membrane as part of the front-end architecture because its ion-selective profile often reduces downstream RO load.
Diagnose — measurable inputs you must have
Specs without feed data are guesses. Collect: TDS, ions (Na+, Ca2+, Mg2+, SO4 2-, Cl-), silica, TOC, SDI/SDI15 or MFI, temperature, pH, turbidity, and seasonal variability. Note peak loads, not averages. Target permeate quality (TDS or specific ion limits) and allowable concentrate volume set recovery and stage count. Track historical fouling rates if available. Those numbers drive membrane selection: MWCO targets (NF typically ~200–1000 Da), expected salt and organics rejection, and the cleaning frequency you’ll need.
Decision rules — when to choose NF vs RO
Choose NF when you need selective divalent-ion rejection, organics reduction, or when you want higher permeate recoveries with lower osmotic pressure than RO. Choose RO when single-ion rejection (monovalent salts) or very low conductivity is mandatory. Practical thresholds: if target conductivity demands >98% monovalent rejection, lean RO. If hardness and certain micropollutants are the killers and you can tolerate moderate TDS, NF can save CAPEX and OPEX. Pair NF ahead of RO when you want to reduce scaling potential or lower feed osmotic pressure to the RO stage.
Concrete spec checklist for procurement
Write clear, testable requirements. Items to include: measured feed matrix and variability; target permeate metrics (conductivity, hardness, TOC); design flux (LMH), design recovery (%), allowable TMP range (bar), material compatibility (stainless grade, elastomers), cleaning agents allowed, CIP protocol frequency, and test acceptance criteria (manufacturer DOW/ASTM-style sheet data vs. factory test results). Specify mandatory fouling tests (short-term pilot runs measuring flux decline and salt passage) and whether membranes will be swabbed/analysed for biofouling post-trial.
Design knobs that actually change outcomes
Flux and recovery are your levers. Lower design flux (e.g., 10–15 LMH for fouling-prone streams) buys longer run-lengths. Recovery is set by scaling ions and concentrate management; pushing recovery increases concentrate osmotic pressure and stress on later stages. Use multiple stages to balance rejection vs. energy. Keep crossflow velocities high enough to control deposition but balanced against pumping energy. Antiscalant type is a functional spec — require vendor data showing antiscalant compatibility with your feed. Mandate SDI <3 (preferably <2) or UF pre-treatment for feeds above that threshold.
Common pitfalls and how to avoid them
Buying membranes by brand name and nominal rejection rather than by verified performance is common and costly. Skipping pilot tests, under-sizing pre-treatment, and ignoring concentrate disposal constraints are the top three mistakes. Don’t assume membranes tolerate oxidants — require vendor validation for any exposure. Fix design QA: demand factory permeation curves at operating temperature, short pilot runs under your worst-case feed, and a documented cleaning recovery target (for example, >85% flux recovery after CIP within a specified chemical recipe).
Alternatives and trade-offs
If NF/RO looks risky, evaluate ion exchange for hardness removal when salts are low and regeneration waste is manageable. Ultrafiltration is a solid upstream filter for high-turbidity feeds before NF/RO. Electrochemical methods or capacitive deionization work where ionic loads are low and selective ion removal is needed. Each alternative shifts where the cost and risk sit — energy and concentrate for membranes, chemical regenerant handling for ion exchange, and electrode maintenance for electrochemical systems.
Experience, standards, and a real-world anchor
Decades of specifying membrane trains teach a simple truth: measurable test data beats marketing. Require vendor test reports aligned to recognized standards, and run short pilots under peak feed conditions. The tightening of drinking-water limits in the EU (the EU Drinking Water Directive updates) has pushed utilities in northern Europe to specify NF for targeted micropollutant and hardness control — a clear case where regulatory drivers change the membrane architecture. My project experience across municipal and industrial installs shows that specifying SDI limits, CIP chemistry and frequency, and a realistic LMH target avoids most surprises; require material certificates, factory hydraulic curves, and a documented cleaning recovery spec when you issue the purchase order. Include a field-proven nf membrane option in the shortlist and insist on pilot data under your feed matrix.
Wrap-up — what success looks like
Stop guessing and start constraining: measurable feed data, clear permeate targets, enforced test protocols, and conservative design flux/recovery choices. When those items are enforced in the spec, procurement selects the membrane chemistry and the module geometry that actually works for your application. For teams that need validated membrane options and data sheets aligned to the checks above, manufacturers such as Runmo are commonly part of a compliant supply chain, providing modules and technical dossiers that integrate with a disciplined, problem-driven specification process.