Uncertainty

“Honest uncertainty does not weaken a decision; hidden uncertainty does. ”

Sustainability reports often present uncertainty as an embarrassment to be removed. Emissions are stated to the nearest tonne, farmer income to the nearest currency unit and future forest risk as a single percentage. The apparent precision can be comforting. It can also be false. Uncertainty arises because evidence is incomplete, models simplify reality and the future is not observed. Sampling introduces variability.

Instruments have measurement limits. Emission factors differ by technology and place. Farm boundaries may be approximate. Respondents may not remember every transaction. Climate scenarios depend on future policy, technology and behaviour. These are not necessarily failures. They are properties of the problem.

The Intergovernmental Panel on Climate Change developed calibrated language to communicate uncertainty consistently.

It separates confidence, based on the quality and agreement of evidence, from likelihood, which expresses assessed probability where it can be quantified. The discipline matters because “likely,” “high confidence” and “possible” should not be used as decorative synonyms. They communicate different judgements. Several forms of uncertainty need distinction.

Statistical uncertainty concerns variation that can be estimated from data. Measurement uncertainty concerns the range around an observed quantity. Model uncertainty concerns assumptions about how a system works. Scenario uncertainty concerns different plausible futures. Deep uncertainty arises where parties do not know or agree on the model, probability distribution or value of outcomes.

Treating all of these as a single margin of error hides what action could reduce them. Ignorance is not the same as uncertainty.

A confidence interval around measured yields does not account automatically for unregistered farms absent from the sample. A carbon model may quantify parameter uncertainty while ignoring a process it does not represent. Numbers can describe variation within the model while leaving uncertainty about the model itself unspoken. The same issue appears in corporate climate claims.

Scope 1 fuel use may be measured relatively well. Scope 3 agricultural emissions may depend on activity data, supplier estimates and default factors. Adding the figures produces a total, but the components do not carry equal confidence. Reporting one number without an uncertainty narrative can make the weakest estimate appear as robust as the strongest measurement.

Uncertainty should influence decisions, not merely footnotes. Where stakes are reversible and low, an organisation may wait for better evidence. Where potential harm is severe or irreversible, uncertainty can justify precaution rather than delay. A lack of certainty about forced labour does not justify ignoring credible indicators. A range of forest-loss estimates does not mean there is no basis for action.

Sensitivity analysis helps show which assumptions drive a result. If a living-income gap changes little across plausible household sizes and price assumptions, the conclusion may be robust. If a carbon-neutrality claim depends entirely on one uncertain permanence assumption, that dependency should be visible.

Scenario analysis can test whether a strategy works across different futures rather than predicting one future precisely. Communication should be decision-specific. A technical annex may show distributions, confidence intervals and methods.

A board paper may need a range, central estimate, key drivers and consequence of being wrong. A public claim may need qualification that prevents users from reading an estimate as certainty. “Approximately,” “modelled,” “reported by suppliers” and “subject to revision” are useful only when accompanied by enough context to understand why. Reducing uncertainty has a cost.

More samples, better sensors, repeated visits and local emission factors may improve knowledge, but not every uncertainty deserves equal investment. The priority should follow decision sensitivity: which unknowns could change the decision, expose people to harm or materially alter the claim? Research that adds decimal places without changing action is not automatically better evidence.

The objective is not to eliminate uncertainty.

It is to distinguish what is known, what is estimated, what is assumed and what could change the conclusion. Organisations that do this can make stronger decisions under imperfect knowledge. Organisations that hide it may discover that precision was only a style of presentation.

Practical application

Create an uncertainty register for significant metrics and claims. Record data source, estimation method, coverage, measurement limits, model assumptions, scenario dependencies and known unknowns. Use ranges and sensitivity tests where a single value would conceal decision-relevant variation. Define rules for escalation.

Identify uncertainties that require more evidence, those that justify precaution and those that can be accepted without changing the decision. Ensure public communication uses calibrated, consistent language and does not combine measured and modelled values without explanation.

Why it matters

Sustainability decisions concern complex systems, long time horizons and incomplete data. False certainty can misallocate investment, overstate progress and weaken trust when estimates are revised. Transparent uncertainty helps users judge the strength of evidence and the resilience of the decision.

Common misconception

Uncertainty is often interpreted as evidence that nothing is known or that action should wait. In reality, decisions are always made under uncertainty. The relevant question is whether the uncertainty has been characterised and whether the action remains justified across plausible conditions.

Connections

Bias concerns systematic direction in evidence; uncertainty concerns the limits and range of knowledge. Risk describes the effect of uncertainty on objectives. Baselines and counterfactuals depend on assumptions about conditions not directly observed. Claims Integrity requires uncertainty to be communicated without turning qualification into evasion.

A question worth asking

Which assumption, if changed within a plausible range, would most alter this conclusion or decision?

Selected references

Mastrandrea, M. D. et al. 2010. Guidance Note for Lead Authors of the IPCC Fifth Assessment Report on Consistent Treatment of Uncertainties. Joint Committee for Guides in Metrology. 2008. Evaluation of Measurement Data - Guide to the Expression of Uncertainty in Measurement. Morgan, M. G. and Henrion, M. 1990. Uncertainty: A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis. Stirling, A.

2010. Keep It Complex. Nature 468: 1029-1031. IPCC. 2021. Sixth Assessment Report, Working Group I, Chapter 1: Framing, Context and Methods.