Oil palm (Elaeis guineensis) has three fruit-forms, high-yielding tenera (thin-shelled), low-yielding dura (thick-shelled) and pisifera (shell-less and female sterile), controlled by any one of nine known mutant alleles of SHELL, a type II MADS-box gene homolog of Arabidopsis SEEDSTICK (STK). We report the development of a non-destructive seed testing approach, which enabled a census of non-tenera contamination via testing of 1,140,603 seed (33.7%) and leaf (66.3%) samples across Malaysia. The prevalence of the high-yielding hybrid tenera form (87.2%) and four undesirable low-yielding contaminant forms (12.8%) were quantified. Removing undesirable palms with SHELL DNA testing prior to field planting would increase fresh fruit bunches and crude palm kernel oil production by 5.6%, while crude palm oil would increase by 7.5% nationally. We present a new paradigm of ‘screen-then-plant’ cultivation using molecular precision agriculture to increase food production on existing plantations, improve sustainability and increase income of commercial sectors and subsistence farmers.