<![CDATA[The reliability of the Black Bunch Census (BBC) for production estimation is influenced by sampling-unit distribution and between-row variability. Precision information is needed to assess stability when census-row composition changes. This study evaluated between-block heterogeneity, between-row variability, mean precision, and census-row adequacy in six mature oil palm blocks in Banyuasin. A total of 1,010 numeric observations from 1,082 palm records were analyzed. Actual zeros were retained, whereas abnormal and dead-palm codes were excluded. The primary inferential units were 34 census rows. Analyses included a row-level Kruskal-Wallis test, Dunn-Holm comparisons, variance decomposition, a 20,000-replicate cluster bootstrap, stability assessment across all observed-row combinations, and age-stratum post-stratification for Block 41B. Between-block differences remained significant at the row level (H(5)=19.440; p=0.0016), although only Blocks 41B and 42B differed after Holm adjustment (p=0.019). Within-row palm variation accounted for 81.47% of total variability, compared with 11.71% between blocks and 6.82% between rows within blocks. The within-block palm-level standard deviation ranged from 1.786 to 2.682 bunches per palm. The relative confidence-interval half-width for Block 41B was 12.65%, while Block 41C was represented by only two rows. Unequal age-stratum sampling in Block 41B reduced the population-weighted mean to 5.895 bunches per palm, 4.67% below the unweighted sample mean. Within these six blocks and one observation period, BBC representativeness depended on row distribution and stratum composition rather than sampled-palm count.]]>
