Abstract Study objectives In-laboratory polysomnography is the current gold standard for objective sleep measurements in clinical trials, but this does not capture night-to-night variability in sleep parameters. This study analyzed variability in sleep parameters recorded over multiple nights of sleep in an ecological setting using a portable sleep-monitor and then estimated the minimum sample sizes required to reliably account for inter- and intra-individual variability in sleep parameters. Methods Participants were males who self-reported the absence of sleep disorders, and used a sleep-monitoring device (Dreem Headband, Dreem, France) over multiple nights of sleep. Night-to-night variability of sleep parameters was determined over five consecutive weeknights using coefficients of variation (CV), and the minimal number of individuals and nights needed to reliably determine each sleep parameter were assessed. Results Night-to-night variability for the whole group (n=94; 470 nights) was high (CV 0.44–0.58) for N2, N3, sleep onset and persistent sleep latencies and wake after sleep onset (WASO), medium (CV 0.22–0.28) for N1 and N3 percentage, awakenings and REM latency, and low (CV 0.04–0.19) for sleep efficiency, N2 and REM percentages, total sleep time (TST) and micro-arousal index. Minimum sample sizes for reliable assessment of TST and WASO were 2 nights with 10 subjects and 4 nights with 50 subjects, respectively. Conclusions Night-to-night variability of sleep parameters is underestimated and underrecognized. These data on variability in commonly-used sleep parameters will facilitate better estimation of sample sizes and number of nights required in clinical trials based on the outcomes of interest.