Recurring financial instabilities have led policymakers to rely on early-warning models to signal financial vulnerabilities. These models rely on ex-post optimization of signaling thresholds on crisis probabilities accounting for preferences between forecast errors, but come with the crucial drawback of unstable thresholds in recursive estimations. We propose two alternatives for threshold setting with similar or better out-of-sample performance; (i) including preferences in the estimation itself and (ii) setting thresholds ex-ante according to preferences only. Given probabilistic model output, it is intuitive that a decision rule is independent of the data or model specification, as thresholds on probabilities represent a willingness to issue a false alarm vis-à-vis missing a crisis. We provide real-world and simulation evidence that this simplification results in stable thresholds, while keeping or improving on out-of-sample performance. Our solution is not restricted to binary-choice models, but directly transferable to the signaling approach and all probabilistic early-warning models.