ABSTRACT Measuring the H i–halo mass scaling relation (HIHM) is fundamental to understanding the role of H i in galaxy formation and its connection to structure formation. While direct measurements of the H i mass in haloes are possible using H i-spectral stacking, the reported shape of the relation depends on the techniques used to measure it (e.g. monotonically increasing with mass versus flat, mass-independent). Using a simulated H i and optical survey produced with the shark semi-analytic galaxy formation model, we investigate how well different observational techniques can recover the intrinsic, theoretically predicted, HIHM relation. We run a galaxy group finder and mimic the H i stacking procedure adopted by different surveys and find we can reproduce their observationally derived HIHM relation. However, none of the adopted techniques recover the underlying HIHM relation predicted by the simulation. We find that systematic effects in halo mass estimates of galaxy groups modify the inferred shape of the HIHM relation from the intrinsic one in the simulation, while contamination by interloping galaxies, not associated with the groups, contribute to the inferred H i mass of a halo mass bin, when using large velocity windows for stacking. The effect of contamination is maximal at $M^{\rm }_{\rm vir}$$∼ 10^{12-12.5}\rm M_{⊙ }$. Stacking methods based on summing the H i emission spectra to infer the mean H i mass of galaxies of different properties belonging to a group suffer minimal contamination but are strongly limited by the use of optical counterparts, which miss the contribution of dwarf galaxies. Deep spectroscopic surveys will provide significant improvements by going deeper while maintaining high spectroscopic completeness; for example, the WAVES survey will recover ∼52 per cent of the total H i mass of the groups with $M^{\rm }_{\rm vir}$ ∼ 1014M⊙ compared to ∼21 per cent in GAMA.