UNDER-SINK RO · WATERSCORE

How we rank and score under-sink RO.

RO-1.2 prioritizes documented treatment evidence, delivery and water efficiency first, then maintenance, ownership cost, installation and warranty. Kitchen RO and whole-house systems are scored separately.

Explore under-sink RO

SEVEN FACTORS · 100 POINTS

What goes into an RO WaterScore.

RO-1.2 uses published editorial weights: 40 points for treatment and evidence, 15 for water efficiency, 15 for daily use and delivery, 10 for maintenance, 10 for five-year equipment and upkeep cost, 5 for installation information, and 5 for warranty/support. Price remains meaningful without outweighing treatment evidence and day-to-day RO performance. WaterScore is not a removal percentage, laboratory result, certification standard, or guarantee that a system makes a particular water supply safe.

40 / 100

Treatment & evidence

Specific contaminant claims, the exact configuration covered and the quality of the supporting evidence. Evidence is incorporated here, not counted again as a separate certification bonus.

10 / 100

Five-year equipment & upkeep cost

Equipment, required accessories, scheduled cartridges and membranes, and mandatory service. Every projection must state its usage, replacement assumptions and exclusions.

15 / 100

Water efficiency

Comparable whole-system water use. Every ratio must identify its direction, operating conditions and which water use is included.

15 / 100

Daily use & delivery

Faucet delivery, useful storage and refill behavior, and sufficient system production for the stated use case. Extra advertised GPD is not unlimited extra credit.

10 / 100

Maintenance

Replacement frequency, required tasks, ease of changing filters and availability of the correct parts. Maintenance cost is evaluated in the cost factor; this factor evaluates the work.

5 / 100

Installation & fit information

Exact dimensions, tank space, service clearance, connections, electrical requirements and a usable model-specific installation manual.

5 / 100

Warranty & support

Written component coverage, duration, exclusions and a clear support and claims process—not simply the largest number in an advertisement.

CLAIM BY CLAIM

The evidence must match the promise.

NSF/ANSI 58 requires a TDS-reduction claim; many other claims are optional. A standard number alone does not prove every contaminant claim. Check the named model, configuration, cartridges and specific claims in the certifier’s listing.

Source: NSF/ANSI 58 requirements and claims ↗

Exact-system certification

Highest evidence credit for the claim actually listed. Accredited certification bodies are evaluated by claim scope, not by which logo appears on the box.

Independent exact-system testing

Substantial credit when a complete report identifies the system, testing method, operating conditions, capacity and results. Testing is not described as ongoing certification.

Documented manufacturer testing

Partial credit for a usable exact-model test record. A manufacturer-posted summary of outside testing is labeled as such when the underlying report is unavailable.

Applicable component or media evidence

Limited credit when the installed component and relevant conditions can be established. Component evidence never becomes a finished-system certification.

Marketing claim only

No evidence credit from an unsupported marketing statement. This describes the limits of the documentation, not proof that the system removes nothing.

FAIR COMPARISONS

Numbers need the same meaning.

Water ratios

A 2:1 clean-to-drain ratio means two treated gallons for one drain gallon. A 2:1 drain-to-clean ratio means the opposite. Under the same measurement boundary, those are approximately 67% and 33% treated water respectively.

We do not silently substitute a membrane recovery figure or an unqualified operating ratio for tested whole-system efficiency. Unclear measurement boundaries stay out of efficiency rankings.

Source: EPA WaterSense definitions ↗

Five-year costs

Start with the exact system and included filters, then count replacements required within the five-year period. Do not count a membrane twice when it is included in a replacement kit. Consider time and rated capacity when the manufacturer specifies both limits.

Installation labor, tax, water, electricity, financing and unplanned repairs are shown separately or explicitly excluded—not silently set to zero. A ten-year figure is a longer projection, not a promise of product life.

Only an exact model verified in the official program listing is described as WaterSense labeled. FilterTier does not award or imply that label.

Source: EPA WaterSense RO guidance ↗

QUALITY AND YOUR FIT ARE DIFFERENT

No automatic bonus for more features.

Stage count, alkaline labeling, mineral cartridges, UV, an app, a screen and a tankless design do not receive automatic overall-score bonuses. Documented function belongs in the relevant factor; preference and installation constraints belong in the household’s fit assessment.

WaterScore

One consistent RO-specific assessment of evidence, cost and ownership tradeoffs. The same model keeps the same WaterScore wherever it appears.

Your fit

A separate match for cabinet space, power, drinking-water demand, treatment needs, budget and preferences. A high overall score cannot override an installation mismatch.

FROM PRODUCT PAGE TO RANKING

A live page does not automatically earn a ranking.

A numeric score requires a reviewed exact-model record, admissible core treatment evidence, all seven factor assessments and source-linked explanations. Missing inputs do not receive invented values, and their weights are not redistributed to make the remaining factors look stronger.

Products with insufficient information can remain in the catalog without a numeric score. Missing documentation is different from a measured failure. Public pages explain material limitations in plain language rather than display internal research statuses.

Overall and format rankings

Our recommendation lists use a stated, category-specific editorial order: exact-model treatment evidence first, then the purchase and replacement tradeoffs and the household use case. The reason for each placement is shown on the page. A rank number is not a WaterScore. When a page explicitly sorts by WaterScore, it uses the same reviewed seven-factor score for that model. Cosmetic faucet finishes do not occupy separate places.

Specialist rankings

PFAS, efficiency and maintenance lists apply documented category requirements and state their sort order. Budget lists verify the actual purchase price. Commercial relationships do not add points or override the published ranking rules.

Discontinued products stay out of rankings. Current but temporarily sold-out products may retain their pages; a buy-now ranking requires current availability. Publish up to ten qualifying choices instead of filling lists with duplicates or unsupported recommendations.

REPEATABLE, NOT PERMANENT

Every score needs an explanation.

A scored product should show its overall WaterScore, factor breakdown, evidence scope, material limitations, methodology version and review date. New certification, specifications, replacement pricing or warranty terms can trigger recalculation.

The method uses documented research unless hands-on or laboratory testing is explicitly identified. Small score differences are not proof of meaningful differences in safety or real-world performance.

Compare under-sink RO systems

PUBLIC SCORE CALIBRATION

Why a strong reviewed system can score in the 90s.

FilterTier keeps two numbers internally. The Research Index is the direct weighted result of the seven factors above. The public WaterScore is a calibrated review score for systems that clear the evidence gate and complete the full audit. This avoids presenting a research index in the 50s or 60s as if it were a failing school grade.

Research Index

The underlying 0–100 weighted calculation. It is retained in the audit record and is used to preserve the spacing and order between products.

Published WaterScore

For a product that passes the scoring gate, WaterScore = 81 + 0.5 × (Research Index − 40), rounded to the nearest whole number and capped at 99. A product below the publish gate receives no WaterScore rather than an inflated number.

The calibration never changes product order. It changes how the same audited performance is presented. A 95 WaterScore does not mean 95% contaminant removal, 95% laboratory performance or a 95% probability of safety.

EDITORIAL RANK IS NOT WATERSCORE

Two different labels. Two clear meanings.

A position such as #1 or #2 is our editorial recommendation within the products and use case named on that guide. A WaterScore is a separate assessment out of 100. We do not convert an editorial position into a numerical score, and a higher position is not a claim that a product removes a higher percentage of contaminants.

First, establish the exact product.

We identify the model, included treatment stages and replacement configuration. A certifier listing is stronger evidence for its named claim than a general statement on a product page. Products with uncertain availability are labeled as alternatives, not confirmed buy-now picks.

Then, explain the tradeoff.

Within the documented shortlist, we consider the applicable claim scope, equipment price, known replacement requirements and the category's purpose. Each recommendation explains why it occupies that position. An absent cost or performance figure is not silently treated as favorable.

Finally, show the limits.

A focused comparison is identified as such. We do not imply that every system in the market was tested or scored. Availability, price conditions, disputed measurements and unverified claim scope remain visible where they affect the buying decision.

The seven WaterScore weights above define the numerical framework. A published numerical score also requires a complete, source-linked factor assessment and consistent calculation rules. The editorial guides below are not sorted by an incomplete or assumed WaterScore.

Explore the RO rankings.

Methodology updated October 6, 2026. This is documentary research, not a claim of hands-on testing by FilterTier. Recommendations and scores can change when the verified evidence changes.