UPDATED 17 NOVEMBER 2026

Water-hardness averages and ranges: which number should you use?

A directory may show one headline value and a range underneath. These are a summary of available evidence, not competing measurements from your kitchen tap. The calculation behind the headline determines what conclusions you can reasonably draw from it.

A midpoint is not necessarily an average

Take a hypothetical published range of 40–160 mg/L as CaCO₃. Its midpoint is (40 + 160) ÷ 2 = 100. That calculation uses only the endpoints. It says nothing about how frequently readings near either endpoint occurred.

Now suppose the actual samples were 40, 40, 40 and 160. Their arithmetic mean is 70, despite having the same range and midpoint. A city page should therefore identify whether its reference estimate is a published mean or a calculated midpoint. These labels explain the evidence; they are not interchangeable wording.

Zone averages introduce another choice

Imagine two zones with means of 50 and 150. Their simple mean is 100. If one supplies nine equal units of water for every one supplied by the other, a volume-weighted mean would be (9 × 50 + 1 × 150) ÷ 10 = 60. Both calculations are valid for their stated inputs, but they answer different summary questions.

Neither calculation identifies the zone serving your address. If you know your zone, its current data is usually the more relevant starting point for an appliance decision. Do not invent weights when the source does not publish them, and do not describe a simple mean of zones as a measured citywide average.

Keep the uncertainty through the conversion

The hypothetical 40–160 range converts to approximately 2.24–8.96 °dH. Converting only its midpoint to 5.60 °dH hides how wide the original range was. If a model’s hardness table puts those endpoints in different rows, the single midpoint does not settle the setting. Follow the manufacturer’s instructions and obtain a relevant tap reading if needed.

Conversely, a narrow range that falls entirely in one manual row may be sufficient for that particular decision. The useful question is whether the uncertainty changes your choice, not whether the converter displays two decimal places.

Not every range is a set of observed extremes

A source may publish minimum and maximum samples, a range across zone averages, or a mean with a standard deviation. These describe different things. A range across zone averages does not contain every individual sample, and an interval calculated from a standard deviation is not an observed minimum and maximum. Read the source’s definition before describing either endpoint as a measured extreme.

For example, a hypothetical mean of 100 with a standard deviation of 10 produces a mean ± three-standard-deviation interval of 70–130. That calculation alone does not tell you whether a sample of 70 or 130 was ever observed. Nor does it establish what proportion of future results will fall inside the interval without assumptions about the data. A page should preserve the calculation label.

Count observations before combining averages

Suppose two groups report means of 50 and 150. If the first contains nine samples and the second contains one, the combined mean of those ten equally weighted samples is 60. If each group instead contains ten samples, the combined mean is 100. The two group means alone do not tell you which calculation applies.

A sample-weighted result describes the collected samples. It is not automatically population-weighted or volume-weighted. A utility may sample a small problem area more often than a large stable area. Without the sampling design and intended weighting, more observations do not necessarily imply a larger share of households or delivered water.

What each summary can tell you
SummaryWhat it representsWhat it does not establish
Range midpointThe point halfway between endpointsThe most common reading
Sample meanThe arithmetic average of the included samplesEvery address in the network
Mean of zone averagesA summary giving each included zone equal weightA volume-weighted network average
Single figureThe one value provided by the sourceZero variation over time

Test whether uncertainty changes the decision

Imagine a fictional manual whose first row applies below 6 °dH and whose second row begins at 6 °dH. A reported range of 80–100 mg/L as CaCO₃ converts to about 4.48–5.60 °dH, entirely within the first row. A range of 80–130 converts to about 4.48–7.28 °dH and crosses the boundary. The second case needs more attention even if both headline estimates appear similar.

This example is not a manufacturer’s setting recommendation. Its purpose is to show how to compare endpoints with a clearly defined boundary. Real manuals may specify their own approach to variation, rounding and overlapping labels. Follow that rule and keep the unrounded conversion until you have located the appropriate row.

Comparing cities without implying false precision

A ranking based on headline figures can be useful for browsing, but differences in sampling period and calculation method limit the comparison. A city at 101 is not meaningfully proven harder than one at 100 when both inputs are approximate. Look at ranges and provenance before interpreting a one-place ranking difference. To make a household decision, the relevant supplier zone matters more than a position in a global list.

Read the basis on a city page

Waterdial records how the estimate was derived, its stated period and source-check date. Those dates have different meanings: checking a report today does not turn its historical measurements into today’s tap samples. A single published figure also does not demonstrate that the supply never varies.

Use the methodology to interpret the basis labels, then consult the linked utility source. The examples above are deliberately hypothetical so you can inspect the arithmetic without mistaking them for newly collected city data. For a household decision, carry the original unit, range and scope into your notes alongside the converted result.

Explore all water and appliance guides · About our editorial approach · Suggest a correction