Multiple methods exist for monitoring changes in vegetation structure, composition and condition over time with method selection representing a trade-off between precision and sampling effort. Visual estimates of canopy cover are commonly used in vegetation monitoring because they are rapid to apply in the field, but these can vary due to subjectivity (i.e. observer bias), differences in observer perspective between height strata and the precision of sampling area relocation during repeat surveys. We used vegetation survey data from shrub swamps in the Blue Mountains to evaluate the extent of variability in the data from each of these sources and the consequences for power to detect changes in cover over time. We also tested if data standardisation would reduce variability due to observer bias and consequently increase power to detect changes. Between-assessor variability was typically higher than within-assessor variability in scores. While variability in cover estimates differed according to growth form, it did not increase with growth form height as expected. Instead, variability was typically higher for growth forms with high cover and vice versa. Contrary to expectation, data standardisation often amplified the differences in scores between assessors, reducing rather than increasing the power to detect change. Small changes in transect location (i.e. 1m displacement) had little effect on cover scores for abundant growth forms but led to discrepancies in cover estimates for rare and patchily distributed growth forms, reducing the power to detect changes over time. We discuss the implications of our findings with respect to monitoring program design.