The requirement for managing an efficient supply chain has always been a balancing act between maintaining high service levels and a healthy inventory turnover ratio. There has been numerous studies and research conducted over the years to address the critical issues facing supply chain practitioners. There have also been many software applications and packages which have been custom-built to ensure that “lost-sales” or “stock outs” do not become a sore point in sales review meetings. This has been mostly done at the expense of low inventory turns and overstocking of parts.
The latest developments in big data technology, which is sweeping across many industries and bringing in huge competitive advantages, can be applied equally reliably to address the challenges faced by supply chain professionals. Big data gives the industry an unprecedented power by bridging both structured and unstructured data and presenting information at the practitioner’s fingertips for quick decision making and insights. The following are some major game-changing rules which big data can bring to the practice of Supply Chain analytics.
1. Leveraging large Volume of Data: A lot of companies have large volume of historical data running into multiple years, or even decades, in some instances. Hadoop’s distributed storage architecture along with compression technologies like Parquet, Avro and ORC enables efficient storage with very fast access. Thus the huge volume of data, which hitherto was not leveraged to its fullest extent, can now be effectively used for advanced analytics.
2. Blending unstructured data for deep intelligence: The availability of NoSQL databases like HBase and Cassandra in the big data landscape enables analytics of unstructured text data which has not been possible until now using legacy Analytics and forecasting packages. This means that information from XML sources for product catalog or web services from suppliers can be integrated in the supply chain decision making process.
3. Advanced analytical models: The Big data community has developed very advanced machine learning algorithms which can be leveraged to used advanced analytical models for forecasting of demand and planning of procurement. Tools like Spark with it’s Machine Learning library (mllib) and R integration in SparkR enable very advanced models to be used on time-series and other data for accurate forecasting and prediction
4. Text analytics: In addition to structured data stored in systems like Hive and semi-structured data stored in HBase, there are numerous tools in the big data toolbox like Elasticsearch and Apache Solr which opens the doors to analyzing text data in various systems. The enormous amount of Textual data can be utilized to gather additional insight about Product feedback, quality and other metrics which can feed into supply chain planning for additional improvements.
5. External data source blending: External data can add a lot of value to demand forecasting or lead time prediction by leveraging real-time information. The advancement in Big Data technologies enables the supply chain software to respond to our ever changing world in a dynamic manner. Hadoop has been successfully used as an ETL tool to unify such disparate data. The data from such external systems can be used to identify potentially new suppliers with better lead times and prices
6. Agility in response: Some of the big data components like Oozie, Sqoop, Flume, Kafka and Storm bring the capabilities of doing procurement in real time rather than periodically. These features makes the company’s supply chain more Agile to respond to a spike in demand, a delay in shipment or a sudden requirement in one of the components in a multi-echelon network.
7. Automated decisions: Gone are the days where supply chain professionals would glean over information in multiple spreadsheets and make procurement decisions. Deep learning systems based on neural networks can now take automated actions based on previously learned data. Moreover these algorithms can get smarter over time by comparing the response against the actual results.