Reduce Time Between Production and Consumption With Agile Machine Learning in CPG
One of the keys to bottom-line profitability in the Consumer Packaged Goods (CPG) industry is efficient inventory turnover. That is, reducing the time between producing goods in a factory to consumption by consumers. Each day the goods stay in the supply chain translates to increase in costs (warehousing, discounting, returns, and disposal). Besides, a mismatch between available inventory and consumer demand leads to missed sales and consumer switching.
For most global CPG brands materials sourcing, product manufacturing, distribution logistics and consumption can occur in different countries. So, sales forecast (“demand sensing”) models from 3, 6, or 12 months ago can miss changes in consumer preferences or in supply chain disruptions. As the past two years has demonstrated, macro and micro forces can bullwhip gluts into shortages and back again faster than the original machine learning models could detect.
Some CPG brands are investing in data science to create refined and granular forecasts for their sales using the power of AI and ML to analyze vastly more data than previously possible. But in our conversation with CTOs, CIOs, and CDOs, we keep hearing the same issue: they are hiring more data scientists than ever and yet they are not seeing the return on investment.
In our experience, the blocker to generating sustained value in AI isn’t the lack of data scientists building accurate ML models. Rather it is the inability of ML tools to keep up with the fast changing environment. When the environment changes (whether it’s an upstream supply chain shock like a container ship blocking a major shipping lane; or downstream demand shock from inflation or fickle consumer preferences), these models quickly degrade. For ML to yield sustainable and substantive returns, CPGs need production ML operations that can adapt as quickly as the environment. This is where Agile ML from Wallaroo can help.
Wallaroo provides all the tooling around deploying, testing, observing and managing ML in production at scale. The agile ML and scalable models can adapt very quickly to the changing environment and hence provide much better accuracy to the future customer demand.
A customer of Wallaroo was testing AI-based dynamic pricing. Initially they started seeing incremental revenue. However, they also found that maintaining the models current to changing market conditions (e.g. a competitor released a new promotion) was nearly impossible for their data science team. Model updates could be built in weeks, but the last mile around testing, validation, and deployment into production could take many months. Additionally, customizing the model for each region required significant resources and time, preventing the data science team from scaling results nationally. Working with Wallaroo they cut time-to-market by over two thirds through simplified deployment and testing. More importantly, they were able to quickly create and deploy ML pipelines for optimized pricing customized for each of their hundreds of regions. By simplifying and accelerating the last mile of ML, they were able to scale their dynamic pricing model from a dozen test locations to thousands of locations nationwide without having to add headcount.
If you are interested to see how Wallaroo can help you unlock the potential of ML to get the right SKU to the right customer via the best, most sustainable distribution channel (3rd party or direct to consumer, online or instore), reach out to us at deployML@wallaroo.ai for a specialist.
