#2 Taking the guesswork out of supply chain forecasting with Machine Learning

Dr. Alexey Drozdetskiy, Head of Data Science for Kainos, explains how machine learning can peel back the layers of complexity in modern supply chains, to increase efficiency, mitigate risk and reduce waste.
Date posted
11 March 2022
Reading time
4 minutes
Alexey Drozdetskiy
Head of Data Science ·

Optimising supply chains with machine learning

Today’s supply chains are complex, global beasts. These networks provide countless business opportunities, but our growing reliance on these intricate webs of relationships and systems means any uncertainty or disruption can have a multiplying effect. With constant changes in the economic landscape, fluctuating supply and demand, and evolving customer habits, it is harder than ever to identify areas of waste and optimisation, accurately forecast demand, or pivot to adjust to changing circumstances.

 

This is where machine learning (ML) is driving real change. ML helps organisations to move away from restricted, rule-based data analysis to create much more refined, specific, and detailed forecasting models. From predicting the impact of changes in the weather, to assessing how much stock will be needed on any given day, ML is helping businesses to take the guesswork out of supply chain management.

 

How machine learning optimises supply chains

Machine learning is a step-up from traditional data analytics. This subset of AI leverages algorithms that process data automatically and detect patterns, mimicking how humans learn and adapting to improve over time.

 

While people can achieve great things with the right datasets, they are constrained by time and can only create predictive models based on limited parameters. At best, this is educated guesswork. ML allows organisations to account for hundreds of parameters, enabling more accurate and precise forecasting. This reflects the multidimensional complexity of real-life situations more precisely and more comprehensively. With limitless memory and compute power, ML can rapidly “crunch” huge volumes of information – including countless forms of unstructured data, visual content, speech, data feeds, and activity logs – far more effectively than humans.

 

This optimises supply chains quickly and continuously, allowing businesses to proactively manage changes in their operating environment – with huge potential benefits.

 

Reducing waste by improving demand projections

The Food and Agriculture Organisation (FAO) of the UN estimated in 2019 that 40% of food waste in the United States was caused by inefficiencies in the supply chain, for reasons including labour shortages, transport and customs issues, and habitual over-ordering. Often this waste can come from poor forecasting. Retailers will choose to stock high levels of popular historical purchase lines to hedge their bets on demand. This produces waste, is not environmentally friendly, and results in extra storage and transport costs.

 

Businesses need demand forecasts to be as accurate as possible to avoid such outcomes. Over-order, and you create waste or incur unnecessary storage costs. Under-order, and you can run out of stock, disappoint customers, and lose revenue.

 

By using ML, retailers can quickly and automatically generate predictive models that are far more accurate. ML algorithms run continuously in the background, providing recommendations and real-time alerts to patterns or changes that need human attention, helping to signal when an intervention may be needed. For example, a food retailer could calculate road conditions, weather, and other factors – like an annual holiday – that might impact supply and demand for perishable goods and adjust their orders accordingly.

 

This type of forecasting and analysis has several applications beyond supply chains. For instance, motor insurance companies can use ML models to predict the impact of weather on claims. A surge in bad weather, for example, could see an insurer shift a greater proportion of the team to incoming calls to handle the influx of claims. Going further, these efforts could also be directed to specific locations if data has been included for contextual variables, such as historically dangerous areas or roads in a state of disrepair.

 

Smarter and more sustainable supply chains

At Kainos, we understand that today’s supply chains must become more sustainable for the benefit of businesses today as well as future generations. Moreover, organisations today have many Environmental, Social and Governance (ESG) requirements that they must meet. Technologies like machine learning can help them achieve their targets in several ways.

 

Machine learning can help businesses meet their sustainability goals by making accurate recommendations for optimising resource management, and transport routes and scheduling. This will reduce the number of needless trucks on the road or containers at sea and make for far more efficiency in supply chains. Not to mention reducing waste of perishable goods that have a high environmental cost to produce.

 

ML can also alert businesses to potential problems in real-time, whether caused by weather conditions, local economic disruption, border issues, signals of modern slavery or labour violations, or as highlighted by the pandemic, illness. Climate change and associated challenges are set to become a challenge in the immediate future, and by using technologies like ML now, businesses can reduce their impact.

 

Building a supply chain for the future

As supply chains continue to feel the strain caused by global events like the pandemic, and socioeconomic events like Brexit, intelligent solutions are no longer an optional asset to have – they are critical to operational success. Kainos provides the expertise to help you start fine-tuning your supply chain. Our focus is on helping you achieve your business goals, streamlining operational efficiency and a providing substantial return on your investment. We were shortlisted in the Outstanding AI ML Project Awards category of Computing AI & Machine Learning Awards for work with clients.

 

If you would like to know more about making machine learning work for your organisation,  contact us today using the form below. 

See how we can help your business

Looking to digitally transform your business? Get in touch to see how we can help you.

About the author

Alexey Drozdetskiy
Head of Data Science ·
Alexey is Head of Data Science at Kainos. He has implemented and led multiple successful projects from Pre-Sales and Advisory to Production and Managed Services for clients ranging from small organisations to FTSE100 companies. He holds a PhD in Particle Physics with a 20+ year career in science and big data. Alexey has authored and co-authored dozens of refereed papers including development of some widely used Machine Learning and statistical algorithms.