This paper introduces a modified principal component technique based on the LASSO (Least Absolute Shrinkage and Selection Operator) to improve the interpretability of principal components. The method incorporates a constraint on the sum of absolute values of the loadings, which can lead to some coefficients being exactly zero, thereby simplifying the interpretation of the components. The technique is compared with traditional principal component analysis (PCA) and rotated principal component analysis (RPCA), and is shown to offer a valuable alternative for exploring the structure of multivariate data. The paper discusses the challenges of interpreting principal components, which often have non-zero coefficients on all variables. Traditional approaches such as rotation are shown to have limitations, including the loss of the successive maximization of variance property. The LASSO-based approach, called SCoTLASS, retains this property while simplifying the components by imposing a constraint on the sum of absolute loadings. This method is demonstrated on the Jeffers pitprop data, where it produces simpler components with fewer non-zero loadings compared to PCA and RPCA. The paper also presents simulation studies showing that SCoTLASS can recover underlying simple structures in datasets more effectively than PCA or RPCA. However, it is noted that SCoTLASS may not always recover the exact underlying structure, especially for certain types of data such as uniform structures. The method is implemented using a projected gradient approach, which is computationally more intensive than PCA but allows for the incorporation of the LASSO constraint. The paper concludes that SCoTLASS offers a useful alternative to traditional PCA and RPCA for simplifying the interpretation of principal components, particularly in cases where the underlying structure is simple. However, there are challenges in determining the optimal value of the tuning parameter and in ensuring that the method finds global optima. Further research is needed to explore the application of SCoTLASS in different contexts and to improve its implementation.This paper introduces a modified principal component technique based on the LASSO (Least Absolute Shrinkage and Selection Operator) to improve the interpretability of principal components. The method incorporates a constraint on the sum of absolute values of the loadings, which can lead to some coefficients being exactly zero, thereby simplifying the interpretation of the components. The technique is compared with traditional principal component analysis (PCA) and rotated principal component analysis (RPCA), and is shown to offer a valuable alternative for exploring the structure of multivariate data. The paper discusses the challenges of interpreting principal components, which often have non-zero coefficients on all variables. Traditional approaches such as rotation are shown to have limitations, including the loss of the successive maximization of variance property. The LASSO-based approach, called SCoTLASS, retains this property while simplifying the components by imposing a constraint on the sum of absolute loadings. This method is demonstrated on the Jeffers pitprop data, where it produces simpler components with fewer non-zero loadings compared to PCA and RPCA. The paper also presents simulation studies showing that SCoTLASS can recover underlying simple structures in datasets more effectively than PCA or RPCA. However, it is noted that SCoTLASS may not always recover the exact underlying structure, especially for certain types of data such as uniform structures. The method is implemented using a projected gradient approach, which is computationally more intensive than PCA but allows for the incorporation of the LASSO constraint. The paper concludes that SCoTLASS offers a useful alternative to traditional PCA and RPCA for simplifying the interpretation of principal components, particularly in cases where the underlying structure is simple. However, there are challenges in determining the optimal value of the tuning parameter and in ensuring that the method finds global optima. Further research is needed to explore the application of SCoTLASS in different contexts and to improve its implementation.