Seismic tomography models show distinct slab structures at different subduction zones and the dominance of degree-2 mantle structure near the core-mantle boundary (CMB). In order to understand how the observed present-day mantle structure came into being, we employ plate-motion history starting from 140 Ma till the present day in global mantle convection models. We tune model parameters such as the effect of different viscosities in the weak layer below 660, weaker asthenosphere and slabs, internal heat generation rate, Clapeyron slope and density change across 660, density and viscosity of thermochemical piles above the CMB, and model duration, to investigate their effect on the predicted mantle structure. We do a quantitative comparison with the predicted present-day mantle structure from our convection models and the seismic tomography models S40RTS and TX2019S, using the long wavelength geoid anomaly as an additional constraint. We find that distinct and linear slab structures are generated when the asthenosphere is moderately weak and stronger slabs are present. The slabs also need to be strong enough since as they sink into the mantle, they sweep the Large Low Shear Velocity provinces (LLSVPs) underneath Africa and the Pacific and generate plumes along their edges. The shape and location of our predicted LLSVPs and slabs are consistent with those in the tomography models. Some of the predicted plume locations are consistent with the real Earth hotspot locations, although the plume structures do not match those in the tomography models. We find that these predicted plumes are integral in fitting the geoid at the intermediate wavelengths. Introducing a moderate Clapeyron slope (-2.5 MPa/K) at 660 does not drastically affect the predicted slab structure. A stronger negative Clapeyron slope resists mass transfer across 660, causing widespread slab stagnation and stalling of plumes at this discontinuity. The predicted present-day mantle structure and the geoid are similar in thermal or thermochemical cases with slightly dense LLSVPs. However, making the LLSVPs highly viscous or more dense hinders plume generation and degrades the fit to the observed geoid. Through this modelling effort, we attempt to constrain the values of various parameters that have an effect on the predicted mantle structure.