E-Land China uses SAS for their decision making system for the Product Planning

For the abundance of big data resulting from its undergoing of a rapid growth — 60% annual average growth — in the last 10 years and for its sophisticated product planning decision-making that reflects the fashion industry, E-Land Fashion China has introduced Big Data solutions, thereby increasing the decision-making accuracies of ordering and distribution by improving and leveling the capabilities of its product planners.

The E-Land Group has undergone an average annual growth of 60% over the past decade in China and it operates 5,241 stores throughout China. The E-Land Fashion China’s sales goal approached 2.5 trillion KRW in 2013. Even though there had been only 2 to 5 E-Land brands in China in the past, in 2013 alone there were over 40 brands including the new brands. As such, a competent product planner -- for ordering and distribution decision-making -- is in high demand for each of the brands. However, the reality is that the rate at which product planners are fostered and trained cannot keep up with the demand. This began the question of, “Why can’t an automated system play the role of product planners to some extent?” And this was the genesis of the Project.

E-land used the Excel since there were no existing analysis systems. Making simulation models for one brand resulted in 26 files with the capacity of these files being whopping 300 MB. In addition, for each brand 110 billion cells were needed for decision-making; where over 15 minutes had to be spent just for a single change in the numbers, and moreover, it was commonplace for these models to be down often. Eventually, a new system was needed, and SAS Big Data solutions were introduced. SAS Enterprise Miner, the SAS Enterprise Guide and the JMP were used to implement an optimized logic for merchandise ordering / distribution, and provided support to enable 30 or so planners can make quick merchandise ordering/distribution simulations and decision-making.

Project Performance and Suggestions

The results were amazing. When a comparison is made -- via simulation –between the actual performance based on 9 brands of the previous summer season without applying the system and the results of the cases when the order quantities and distribution quantities changed with the system applied, it was discovered that a 29.1 billion KRW improvement in operating profit (75% growth compared to previous year), which is in a way a planning loss. Looking at this in detail, the brand-specific strategic product concentration levels (the order proportion for the top 20% styles) could have been increased from 38% to 47%. Furthermore, by reducing the region-specific TCR variations, the sales quantity of 0.386 million pieces could have been increased to 1.102 million pieces, and in actuality, the error rate for the product mix (assortment) sales rate was reduced from 25% to 2%.

In the meantime, E-land would like to make the following suggestions to those who are about to proceed with similar projects:

• First, it is important to reset the data analysis and decision-making units in order to automate the system. The standardization of the reference information would be included in this.
• Second, it is important to have a systems point of view. Rather than looking at the system as a “system that guarantees success,” it would be more desirable to look at it as “something that prevents failure.” As such, it is prudent to have the decision-making offered by the system be a “range” rather than a specific value. In other words, it is recommended that the system provide a decision range within which a person can make a decision, i.e., a Decision Support System.
• Third, although improvement in the prediction accuracy is important, in actuality, “reduction in the prediction period” is more important. For this, it would be more effective to improve the processing speed at which products are made, in conjunction.
• Lastly, the “value validation” must be repeated indefinitely until it is a perfect fit. Since risks can be prevented only by verification as the pricing for decision-making is higher, we urge you to spend as much time and effort in value validation as in the development.