Find the fault.
Then fix it.

Alkalinity falls between doses faster than expected, confirmed by trend testing in an intervention-free window. 33-step fault tree: test confidence, trend confounds, supply and delivery verification, consumption vs. precipitation.

Nitrate reads above the applicable upper boundary for your tank type, confirmed by a second method. 26-step fault tree: reading confirmation, carbon-dosing and phosphate gates, tank-specific critical classification, preflighted calculated water change, then source, export, and input isolation.

Nitrate reads undetectable, or repeatedly bottoms out, confirmed on a second method. 19-step fault tree: zero and persistence confirmation, cyanobacteria attribution gate, carbon-dosing documentation gate, then export and nitrogen-input isolation. Corrections restore input or reduce export one measured step at a time: nothing is dosed from a bottle.

Phosphate reads above the applicable upper boundary for your tank type and keeps climbing, confirmed by a second method. 28-step fault tree: reading confirmation, nitrate classification, chemical-removal method boundary, tank-specific critical classification, preflighted calculated water change, then source, export, and input isolation.

Phosphate reads undetectable, below the selected method's reporting capability, or repeatedly bottoms out after previously measurable results. 37-step fault tree: method qualification, fresh-sample confirmation, nitrate classification, then five separate removable-cause branches (binding media, lanthanum, carbon dosing, refugium or scrubber export, and input restoration). A confirmed high-nitrate imbalance walks its own route and never enters feeding restoration. Nothing is dosed from a bottle.

Dosed calcium fails to rise, or keeps falling, across comparable tests despite an unchanged dosing program. 32-step fault tree: reading and trend confirmation, confound window, magnesium prerequisite, supply and delivery verification, then consumption corroboration.

Alkalinity climbs from test to test while the dosing program sits unchanged. 27-step fault tree: reading and trend confirmation, confound window, kalkwasser and dosing-input checks, delivery verification, then demand corroboration. Every correction reduces an input: nothing is dosed to force the level down.

Magnesium falls across comparable tests spaced days apart despite stable husbandry. 28-step fault tree: reading and trend confirmation, confound window, above-range disposition, supply and delivery verification with a salt-mix check, then calcification corroboration.

pH reads below the applicable lower boundary at the same time of day on three or more days. 37-step fault tree: meter qualification, high-side exclusion, fresh-reading reproduction, two-sided fresh alkalinity classification, an established kalkwasser program's integrity check, then a three-outcome aeration comparison routing to room-air or aquarium-side causes. Every correction restores something that already existed: nothing is purchased, no new program is started, and no dose is invented.

A salinity value, from an ICP report or any home reading, sits outside the acceptable band for the tank type. 28-step fault tree: instrument qualification with a seawater-range standard check, fresh-reading reproduction, severe-deviation screening, direction classification, then top-off and water-source cause isolation. Corrections are a bounded evaporation-substitution raise or a calculated dilution (never dry salt in the tank, never an invented volume), each verified over seven days of same-time readings.