Background: Early cannulation arteriovenous grafts (ecAVG) for dialysis access are limited by reintervention for venous stenosis (VS) despite their good initial patency. Whilst stent-grafts (SG) have shown promise, the optimal sizing is unclear. Therefore, this study aims to determine if outflow vein diameter, SG diameter or these relative to each other (V:Sr) alters outcomes, and if so, which is more important. Methods: Retrospective analysis was performed of Gore® Acuseal® ecAVGs with VS treated with Gore® Viabahn® SG over a 7-year period. Primary patency (PP), time to thrombosis and functional patency were analysed by SG length/diameter, vein diameter and V:Sr. Results: We identified 114 ecAVGs with median follow-up 492 days (IQR 189–770). SG length and diameter did not correlate with PP, however, there was a significant relationship between vein diameter and PP (RR = 0.901 (0.832–0.975), p = 0.01) and between V:Sr and PP (RR = 0.462 (0.255–0.838), x2 = 5.866, p = 0.0015). The optimal V:Sr was ⩾1.4 (i.e. vein diameter at least 40% greater than the stent-graft; or ‘free-floating’ stent outflow) (RR = 2.759 (1.670–4.558), p < 0.001), translating to a difference in median PP of 252 versus 496 days (IQR: 188–316; 322–670). On multivariate analysis, absolute vein diameter lost significance, whilst V:Sr remained an independently significant predictor of PP (RR = 3.247 (1.560–6.759), p = 0.02). Conclusions: Placement of the SG outflow into a relatively larger segment of vein was associated with a significant increase in PP independent of the absolute vein diameter. This suggests that larger calibre SG which are apposed to the vein wall are not required for optimal primary patency, and indeed should be actively avoided. Instead, a ‘free-floating’ stent outflow which is undersized relative to the recipient vein (whilst maintaining a minimum anchoring calibre) is recommended where possible. This should be considered during intervention and may require selection of longer devices, where practical, to bring the stent outflow into a larger vein segment. Level of Evidence: Level 3a, Non-randomised controlled cohort/follow-up study