Since the discovery of the accelerating expansion of the Universe more than two decades ago, Type Ia Supernovae (SNe~Ia) have been extensively used as standardisable candles in the optical. However, SNe~Ia have shown to be more homogeneous in the near-infrared (NIR), where the effect of dust extinction is also attenuated. In this work, we explore the possibility of using a low number of NIR observations for accurate distance estimations, given the homogeneity at these wavelengths. We found that one epoch in J and/or H band, plus good extit{gr}-band coverage, gives an accurate estimation of peak magnitudes in the J (Jmax) and H (Hmax) bands. The use of a single NIR epoch only introduces an additional scatter of ~$0.05$ mag for epochs around the time of extit{B}-band peak magnitude (Tmax). We also tested the effect of optical cadence and signal-to-noise ratio (S/N) in the estimation of Tmax and its uncertainty propagation to the NIR peak magnitudes. Both cadence and S/N have a similar contribution, where we constrained the introduced scatter of each to $<0.02$ mag in Jmax and $<0.01$ in Hmax. However, these effects are expected to be negligible, provided the data quality is comparable to that obtained for observations of nearby SNe ($z < 0.1$). The effect of S/N in the NIR was tested as well. For SNe~Ia at $0.08 < z < 0.1$, NIR observations with better S/N than that found in the CSP sample is necessary to constrain the introduced scatter to a minimum ($< 0.05$ mag). These results provide confidence for our FLOWS project that is aimed at using SNe~Ia with public ZTF optical light curves and few NIR epochs to map out the peculiar velocity field of the local Universe. This will allow us to determine the distribution of dark matter in our own supercluster, Laniakea, and to test the standard cosmological model by measuring the growth rate of structures, parameterised by $fD$, and the Hubble-Lemaître constant, $H_0$. All of the software developed and used throughout this work is publicly available.