How signal processing can optimize non-seasonal product inventory

Winston owns and operates Winston’s Wristwatch Warehouse, a successful chain of stores dedicated to things that tick. They offer tens of thousands of high-end and mid-range watches from hundreds of international vendors. While Winston’s brand has been running like clockwork for years, he has noticed that several local competitors have begun to challenge his business. These rival watch merchants are now cutting into Winston’s share of the market.

How can Winston help his business prevail without sacrificing the quality and customer service that earned him his reputation?

Keeping your inventory lean

Winston’s answer lies in minimizing operating costs by optimizing his inventory. Most businesses severely limit their profitability by either spending too much on inventory or not stocking enough products to keep up with demand. The least successful among them are the ones that “follow their gut” and replenish their stock in an arbitrary and haphazard fashion. Others use previous years’ sales as a guide for reordering products. Even more statistically savvy businesses perform time-series analysis to estimate how many of each item to buy.

However, it’s no simple matter to accurately forecast changes in demand at each store for the thousands of products stocked by businesses like Winston’s. In this particular example, Winston’s flagship products are big-ticket items that cost thousands of dollars, have long lifespans, and lack a clear peak purchasing season. These characteristics make Winston’s business a poor fit for time-series analysis.

Worry not: digital signal processing can break this kind of complex pattern into simpler component patterns. Using signal processing for deep forecasting, data scientists can forecast demand with 60 percent less error than traditional statistical techniques.

Conclusion

Signal processing is a powerful technique that can be used in situations that aren’t suitable for more common time-series analysis. Combined with machine learning, this method can yield excellent accuracy for each product and store, even when sales data is incomplete.

To learn more about how Visionet’s inventory planning solution generates accurate demand forecasts using signal processing and machine learning, please download our white paper, “How to Slash Inventory Using Signal Processing & Machine Learning”.

Download White Paper

Striking gold with AcuitySpark

“Retail is detail.”

Chadatip Chutrakul, CEO of Siam Piwat, Thailand Thousands of customers walk in and out of your store or visit your website every single day. Do you know their likes and dislikes? What are their preferences in terms of products, colors, and styles? If you want to upsell or cross-sell products to them, do you know where to direct them? Are you aware of their buying power, their lifetime value, and whether or not they're likely to churn?

A clean, integrated customer dataset is a goldmine for retailers looking to improve marketing ROI and market share, and is key for retailers wanting to stay ahead of their competition. Retailers usually struggle to maintain what has popularly become known as a “Golden Record”. Studies have revealed that only 30 to 40 percent of customers are actually known to retailers unless a proper strategy is established and the necessary tools are put in place to maintain customer master data and convert unknown customers into known ones.

Even companies that have a strong CRM framework struggle to build a Golden Record because:

• Few attributes are typically available for unknown customers. This hinders the construction of a complete customer persona.
• In eCommerce, the segment of unknown customers is large due to the availability of the guest checkout option. As a result, every transaction is treated as one made by a new customer.

The figure below illustrates this problem graphically.

Given the hefty benefits of knowing your customer, customer relationship and marketing departments devise methods to identify customers and track their activity and transactions over time so they can move as many people as possible from quadrants 1 and 2 to quadrants 3 and 4. These methods include giving out loyalty cards, offering discounts on social media registration, and encouraging people to make purchases via mobile apps.

While these methods have their merits, they disregard historical data and transactions, and only work for customers that buy into these programs. A customer merge process can overcome both of these issues and create a Golden Record using the following method:
 A properly executed customer merge process can yield a significant improvement in the percentage of customers that are known.

 

AcuitySpark, Visionet’s proprietary solution for retail analytics, has a prebuilt module for merging customers that has been developed using top-of-the-line practices, tools, and technologies. Our clients have been able to attain Golden Records and move up to 45% percent of existing customers from quadrants 1 and 2 to 3 and 4 with minimum effort. Interested? Get in touch with Visionet today.