Sensitivity analysis: which assumption is actually carrying the result?

A life cycle assessment rests on dozens of assumptions: transport distances, electricity mix, service lives, recycling rates, allocation choices. They are not equally important. Sensitivity analysis shows which ones actually carry the result.

Sensitivity and uncertainty are not the same

Uncertainty analysis asks how much spread known input distributions produce in the output; Monte Carlo simulation is its typical tool. Sensitivity analysis asks a different question: how much does the result move if I change this parameter? The first measures uncertainty, the second finds the levers.

The two are complementary, but sensitivity analysis is usually more useful because it suggests an action directly: improve the data behind the parameter that carries the result.

One-at-a-time scanning

The simplest and most common method is changing each parameter individually by a set proportion — typically ten or twenty per cent — and measuring the resulting change. The output is a list ranking parameters by the size of their effect, and that list sets your data collection priorities directly.

• Electricity emission factor — almost always first for energy-intensive processes
• Main raw material quantity and source
• Transport distance — decisive for low-density or heavy products
• Product life and replacement frequency — critical in whole-life calculations
• Recycling rate — large effect in Module D and under the cut-off approach
• Allocation method — can reverse the result on its own in multi-output systems

Mandatory in comparative studies

In a study comparing two products, sensitivity analysis is not optional. If the difference in the result is ten per cent while a critical parameter moves thirty per cent across its plausible range, the comparison is meaningless. ISO 14044 requires such checks for comparative assertions disclosed to the public.

The practical rule: if you cannot show the difference exceeds the swing from the most sensitive parameter's plausible range, it is more honest to say the two products are equivalent.