The bonding that exists between the old concrete and the new concrete depends largely on the quality of substrate surface preparation. The accurate representation of substrate surface roughness can help determine very precisely the correct bonding behavior. In this work, an experimental investigation was carried out to quantify the normal concrete (NC) substrate roughness parameters and evaluate their relationship with the bonding performance of ultra high-performance fiber concrete (UHPFC), used as a repair material. The bond strength was quantified based on the results of the pull-off test, splitting cylinder tensile test, and the slant shear test. Three types of NC substrate surface preparation were used: as-cast (without surface preparation) as reference, wire-brushed, and sand-blasted (SB); the roughness of which was determined using an optical three-dimensional (3D) surface metrology device (Alicona Infinite Focus). It was observed from the result of the pull-off test that failure occurred in the substrate, even though adequate substrate surface roughness was provided. Moreover, analysis of the splitting cylinder tensile and slant shear test results showed that the substrate surface preparation method had a significant influence in bonding strength between UHPFC and the NC substrate. The composite UHPFC/NC substrate having a SB surface behaved closely as a monolithic structure under splitting and slant shear tests. An excellent correlation (R-2>85%) was obtained between the substrate roughness parameters and the results of the splitting cylinder tensile and slant shear tests.