Supplier Planning: It’s all about Big Data

In AI, Big Data, Machine learning by VendorMach

Over a fifth of 400 enterprise buyers blamed supplier failure for higher insurance premiums, damaged reputation and loss of customer trust according to a Tungsten report.

The challenges faced by businesses include cyber fraud and siloed data which impact customer service, product quality and operations.

Is it really all about the data?

Yes. Data is growing faster than ever before and by the year 2020, about 1.7 megabytes of new information will be created every second for every human being on the planet.

IDC estimates that unstructured content already accounts for a staggering 90 percent of all digital data, much of which is locked away across a variety of different data stores, in different locations and in varying formats.

Poor integration of systems and limited data sharing within the supply chain product flow cost retail companies $158.5 billion according to RetailWire. Operational challenges include disruptive effects – suppliers delivering at the wrong time or not being able to fill orders impacting sales.

More planning excel spreadsheets and pivot tables?

There is clearly an opportunity to leverage data as an asset, not a burden. Every planning spreadsheet a team creates feeds into more megabytes of silo-ed information that is locked away without any insights. Here are the some of the issues with spreadsheets

  • Error-prone due to its manual nature.
  • Only accessible locally.
  • Challenge to merge input from multiple planning and performance teams.
  • Impossible to consolidate input of data and limited insights of historic and or third party data.
  • Limited comparisons of suppliers and their data without substantial manual effort including substantial rework – no version control.
  • Challenge to track supplier KPIs in real-time.

Next Gen: Machine learning to the rescue?

Forbes argues that machine learning (ML) is a hot topic, but it’s not new for supply chain management particularly demand forecasting. The same can be argued for the insurance industry (pricing coverage based on longitudinal population data for decades) or the utilities industry (pricing meter and longitudinal usage data). It is important to note that time series analysis and forecasts based on traditional structured data is not machine learning. Solid approaches to ML ensures that structured, semi structured and un-structured data collection across both internal and external parties is optimal, is predictive (supervised or unsupervised training models occur over a period of time), changes real-time based on new information, is conclusive enough to support staff decision making, addresses security compliance and has transformative enterprise impacts.