Retailers cut down ‘out-of-stock’ using lost sales forecasting

About client

Client is a clothing manufacturer and retailer with eCommerce platforms as predominant sales channels. They sell a wide variety of women's clothing items and have distributed warehouses to support their fulfillment operations. Noticing the dynamically changing demand and trends for their products, they wanted to do in-depth forecasting to support their business decision-making and to identify lost-sales.

Challenges

Client’s main goal is to bring demand and sales aligned in one direction and improve revenue opportunities. Some of the challenges they faced since they didn’t get supportive insights from data include:

Lost sales opportunities: Selling various products in different colors and sizes on multiple platforms like Myntra, Flipkart, and Amazon, they are unable to keep up with changing demand. This often leads to lost sales opportunities where their products get stocked out on one or many channels. It will take a few more business days for them to stock them back up, allowing their competitors to seize the sales opportunities.

No insights available for prediction: Retail manufacturers with eCommerce sales must know their demand ahead to maintain the right level of stocks on both high and detailed levels. They wanted to know which product range of which size and color will be in demand in which regions. This could help them run inventory and marketing operations smoothly, increasing focus on what’s more in demand. 

For example, running platform-specific ads using demand prediction insights which will increase both marketing ROI and product sales.

No centralized records: They do not have any centralized data repository or analytics systems that they could refer to. It’s only their sales and order invoicing system and marketing advertisements tracking applications operating in silos. This hindered their view on customer analytics as well as product sales trends. They had to rely on manual data analysis and guess-based decision making.

Skewed analysis due to manual entry: The siloed data systems caused many lags and delays in the data entry, leading to inaccurate ROI measurements and analysis.

datakulture solution

Our data science team started with exploratory data analysis. We aggregated their sales and marketing data to understand periodic sales trends for different products and channels. 

Major discovery: We conducted Pareto analysis, which is finding their most profitable product categories and their respective eCommerce sales channels. 

Sales forecasting model: Collecting their sales and marketing data, we did data processing, cleansing, and transformation, and moved data to a data mart for data storage. Built a base sales forecasting model using the FB-Prophet model, which generated forecasting reports for the next three months. 

Mitigate lost sales: Based on the inventory levels, the model predicted the possible stock-out instances across different eCommerce platforms. Along with this, we shared reports with recommendations and to-do’s on avoiding these lost-sales opportunities.

Visualizing reports: Develop a set of Power BI reports to demonstrate the lost sales and forecasting trends. These reports are aimed at different levels of business users and will aid them with quick decision making. 

Included the option to perform ‘What-if analysis’, which helped them adjust values in reports and see ROI ahead of time, something that could help them set distinct goals.

Conclusion

Our first level study of their data helped us convey the top three product categories that contribute 85% of their sales and which channels contributed the most. Besides, the biweekly data forecasting report contained the drilled-down view of data, including what line of products sell in what quantities and specifications. All these insights, along with visualized reports, helped them pick up and respond to market shifts fast, and maintain ideal inventory levels across their warehouses. They could tune their advertising and marketing campaigns better and make the most of the changing demand.