Supply Chain Analytics

Agile demand sensing model drives higher forecast accuracy across products for global CPG company

» Agile demand sensing model drives higher forecast accuracy across products for global CPG company

A UK-based multinational consumer goods company delivers health, hygiene and nutrition products to a global clientele through a complex supply chain network of 40,000+ retailers. They have also partnered with an ecommerce giant to push sales across product lines.

Business Need

The ecommerce company’s SLA mandated adequate inventory stocking at any point of time and failure to do so would invite heavy penalties.

Legacy systems threw up inaccurate demand and sales forecasts and challenged the company’s ability to respond to fluctuating market conditions. COVID further disrupted stock availability, and adversely impacted the consumer goods company’s customer experience and online sales.

The consumer goods company hence partnered with Hitech to create an intelligent demand sensing and sales forecasting solution to:

Set in place agile forecasting systems to discern historical data from market segmentation, geographic and demographic factors
Generate accurate datasets to feed into algorithms

Want to get integrated and real-time visibility across your supply chain operations?

Get in Touch with us »

Solution

The team at Hitech Analytics designed Machine Learning-based demand sensing solutions to process voluminous data, provide algorithmic analysis and deliver intelligent supply chain related insights.

The solution focused on:

Collecting data from customer feedback, behavior patterns, and sentiments across geographies
Executing different regression algorithms for data validation
Optimizing the algorithm to predict demand in near future for all products
Improvising demand sensing model to match changing market requirements

Approach

In the initial phase, our data specialists interacted with the company’s product and forecast managers to identify the actual gaps in their existing analysis and data models. They concluded that:

In a normal market scenario, their existing time series algorithm could use data from parameters like seasonality, cyclicity, and standardized patterns of sales data.
It failed with dynamic inputs like fluctuation in online sales, marketing promotions, disruption in local environment (Covid-19 in this scenario), etc.

The client shared huge datasets on total inventory, number of daily site visits, number of people visiting the company’s portal on the ecommerce site, customer reviews on individual products, etc.

Our team changed the time series approach to a regression one. As algorithms were trained on specific datasets, we also changed existing parameters to lag parameters for data analysis. It helped us to establish a correlation between time duration and parameter’s impact.

We used Machine learning and regression algorithms like XG Boost and decision for accurate forecasts for shorter time periods.

Software & Technology Used

Python, Excel and ML-algorithms

Results

Using a cognitive demand sensing solution for various markets, the consumer goods company was able to:

Build a resilient and adaptable supply chain
Get more accurate real-time forecasts at product-mix level
Ensure on-time delivery for improved customer experience
Save on huge penalties and lost sales
Improve working capital by maintaining right inventory levels