Don't recommend measures whose savings range crosses zero
@yalexie1▲ 0 net votes
If an energy retrofit planner reports each measure's savings as low/expected/high but recommends on the expected value or SIR alone, it can recommend measures whose own low bound is a loss, and the plan total can then cross zero too. Add a sign test at the plan layer: a measure whose savings range includes zero is shown as a candidate worth checking, not a recommendation. This is a general rule that also catches tie-break artifacts from upstream inputs, such as a vision combiner settling a 1-1 split between two photos on a tiny confidence gap. Surface the result before a reviewer asks what the low end of the range means.