1. Abundance trends provide key guidance when setting conservation priorities, whether indicating population decline, stability or recovery. Knowledge of the power of surveys to detect trends is essential, as the consequences of not detecting a real trend can be profound. 2. Unfortunately, some surveys have been established with no assessment of power, and others are used to study species that were not their original focus. The latter is common in the marine environment, where rare fish are monitored using catch data from surveys that target more abundant commercially fished species. 3. We calculated the power of a large-scale annual monitoring survey (the English North Sea bottom trawl survey) to detect decline and recovery of species that are vulnerable to fishing. As fisheries exploitation invariably precedes scientific investigation, the survey began after many vulnerable species had already been depleted. 4. The power of the survey to detect declines in the abundance of vulnerable species on time scales of < 10 years was low and the survey often failed to detect declines that would lead to listings under the IUCN A1 Red List criteria. Thus conservation prioritization based solely on survey data may fail to identify species at risk of regional extinction. 5. If conservation measures were effective, and vulnerable populations recovered at the maximum potential rate, 5-10 years of monitoring would often be required to detect recovery. 6. Power to detect trends in abundance was increased by developing a composite indicator that reflected trends in abundance of several vulnerable species. This indicator provided an overview of their conservation status. 7. Synthesis and applications. Consistent with the precautionary principle, conservation prioritization and management action should not depend on the statistical significance of recent abundance trends when low power is a consequence of historical depletion. If the conservation prioritization and management of rare and/or vulnerable species have to be predicated on evidence of significant declines, then higher type 1 error rates (falsely detecting a decline) should be acceptable. This is because the costs of type 1 errors are lower than those of type 2 (failure to detect a real decline).