The objectives of the current study were to determine the variation structure within a day and across days when determining dry matter intake (DMI) of Coastal bermudagrass (Cynodon dactylon (L.) Pers.) pasture using dotriacontane (C32) as an external marker, to determine the optimal times for faecal collection for forage DMI estimation and to compare C31 and C33 as internal markers in estimating forage DMI in Brahman bulls. Sixteen Brahman bulls were allocated by weight to four pastures, and stocked at a moderate to low grazing pressure for 63 days from late June to the end of August. Three intake measurement periods (P1, P2, P3) were used; each period consisting of 10 days of twice daily C32 (400 mg/day) administration. Faecal collections were taken during the last 5 days (07.00, 11.00, 15.00 and 19.00 h). The C32 was individually hand fed using Calan gates, with maize gluten as a carrier for the alkane. Gas chromatography was used to determine n-alkanes in the forage and faecal samples. The concentration of C31 was less than C33 in the bermudagrass for all periods (P < 0·001), but the concentration of C31 and C33 in faeces was not significantly different (P > 0·05). The average concentration of C32 in the forage was 5·1, 7·6 and 9·6 mg/kg dry matter (DM), for P1, P2 and P3, respectively, with an average of 7·5 mg/kg DM for all periods. During P1 and P2, the estimation of forage DMI using C33 had a better fit (smaller –2 × log and Akaike's information criterion (AIC)) than using C31 either with or without adjustments for C32. The variation in estimated forage DMI decreased when forage C32 was not included. The variances of forage DMI were similar using C31 across days, but the Pearson correlations between days were low, which suggested that several days of collection were needed to estimate forage DMI accurately. Correlations between collection times within days were medium to high for all periods and varied from 0·65 to 0.97 for C31 and from 0·26 to 0·96 for C33. When all periods were analysed together, estimates of forage DMI either using C31 or C33 had low correlations between days of collection. Adjustment for C32 did not improve the variance and (co)variance matrix. In summary, C33/C32 had the lowest variation in estimating forage DMI, and at least 5 days of faecal collection were needed to decrease the variability of estimating forage DMI. The optimum times for faecal collection were 07.00 and 19.00 h, and it was important to adjust for C32 alkane concentration in estimating forage DMI in Brahman bulls grazing Coastal bermudagrass.