In the oil and gas industry, the current standard that is used to quantify the fraction of water (so called “water-cut”) in an oil–water multiphase flow stipulates the oil–water mixture to be homogenised to ensure sampling is representative. Although there are devices that comply with the minimum requirements of the sindustry standards for custody transfer applications, our understanding of the homogenisation process is limited; where even small errors arising due to inhomogeneity could cost tens of millions of dollars annually per metering station. To that end, we have developed a flow loop and homogenisation process to study oil–water multiphase flow. Experimental investigations were carried out using magnetic resonance (MR) imaging and hence the entire flow loop has been designed to fit within a MR laboratory, with the homogenisation step itself performed within the bore of the magnet. Measurements were performed in a 2.5″ diameter Perspex pipe at stream velocities between 0.2 ms$\mathrm{<mrow\; data-mjx-texclass="ORD">-1</mrow>}$ and 1.47 ms$\mathrm{<mrow\; data-mjx-texclass="ORD">-1</mrow>}$, to mimic typical pipeline conditions. The size of the pipe diameter used in this study is unique compared to previous studies for oil–water flow applications using MR. To facilitate experimental investigation, we have developed MR techniques to quantify the water-cut and improve our understanding of mixing in liquid–liquid flows. Chemical shift selective (CHESS) MR was used to quantify the water-cut between 2.5% and 25% for static samples. These results show a linear relationship and demonstrate that the water cut is measured with an accuracy of ±0.2%. The CHESS sequence was combined with MR imaging sequences to enable visualisation of the water distribution in real time in one-dimension, or as a time-averaged measurement in two dimensions. MR measurements were also performed on an oil–water multiphase flow at a stream velocity of 0.2 ms$\mathrm{<mrow\; data-mjx-texclass="ORD">-1</mrow>}$ and for water cuts between 1% and 7.5%. Local measurements of the water cut are performed with an error of less than 1%.