This study aims to optimize the K-Means algorithm to improve the clustering model of fashion goods sales transaction data at the josi.id store over a period of seven months. One of the main challenges is the lack of understanding of the characteristics of sales transaction data at the josi.id store, as well as the difficulty in identifying products that cause spikes on big days. With the K-Means clustering method used to group data, the optimal K value, attributes that affect the Davies Bouldin index (DBI) value. The analysis of the results shows that the key attribute that affects the k value is the TYPE OF ITEM with K = 3 as the optimal value, has the lowest DBI value of 0.258 compared to other cluster configurations. With the characteristics of cluster 0 (429 items) showing dominant sales during the Eid season. Cluster 1 (343 items) shows high sales during the holiday period. Cluster 2 (309 items) has stable sales during weekdays. These results show good separation and uniformity of clusters in each cluster. The attribute of ITEM TYPE, based on the characteristics of each cluster is Bracket clothes products show the highest total sales of up to 7 million, supported by traffic (love feature) that is often viewed. Blouses have total sales of under 2 million, while dresses show great variation with total sales between 1 and more than 3 million. Skirts have a more diverse sales distribution, with transactions reaching 3 million. which includes categories such as Dresses, bracket clothes, Tops, and Skirts, plays an important role in grouping sales transaction data, especially for seasonal products such as during Eid.