This website uses cookies primarily for visitor analytics. Certain pages will ask you to fill in contact details to receive additional information. On these pages you have the option of having the site log your details for future visits. Indicating you want the site to remember your details will place a cookie on your device. To view our full cookie policy, please click here. You can also view it at any time by going to our Contact Us page.

Plotting demand for consumer-packaged goods

30 April 2021

Richard Seel argues that historical forecasting has had its day across the entire consumer-packaged goods sector. 

Today, historical forecasting seems to have very little value. The advent of the Covid-19 pandemic reset all the parameters. It would have been nonsensical, for example, to attempt to predict the demand for toilet roll and sanitary products in March 2021, based on patterns seen 12 months before. Likewise, as winter set in, supermarkets preparing for Christmas would have been well advised not to put too much store on last year’s figures when planning for the festive season. 
A data-driven solution
Food manufacturing businesses need to quickly adopt a more agile, data-driven approach. In times of uncertainty, a new kind of predictive modelling is called for, typically incorporating, (or being accompanied by) an element of scenario planning. Could switching to predictive forecasting and modelling offer a solution?
This new approach enables food manufacturers and their supply chain partners to understand how the current market is evolving so they can forecast the future, even where they have little or known disrupted historical insight. The increased complexity of the approach reflects the difficult scenarios organisations are having to wrestle with in the Covid era. We are seeing a paradigm shift in the way forecasting is undertaken. Thanks to the lack of reliable historical data, an organisation’s entire approach has moved away from a traditional time series focus on ‘when’ certain events are likely to happen, to a more classification-focused or predictive model - likely to change in real-time - that is designed to pinpoint the probability that something will happen at all.     
In line with this, we are seeing an inflation of variables influencing the model. Rather than relying purely on internal models, food manufacturers are looking to a growing array of third-party sources to get the detail they need – everything from social media to governmental databases, and from business information websites to medical and financial sources. Data from a range of sources helps organisations to build better forecasting accuracy and reliability in what is an increasingly uncertain environment.    
This is still an emerging field today. In the short-term at least, a hybrid approach will likely be adopted. Different solution sets, including cloud and predictive analytics, are gradually integrating. For the time being, organisations will need to be creative in tapping into different data streams and bringing their insights together. However, the market for these kinds of predictive forecasting and scenario modelling solutions is likely to grow significantly over time, with those organisations that have the right tools in place and the right levels of maturity, in terms of analytics, leading the way. 
Over the longer term, there are two approaches that manufacturers will need to get right if they are to make a success of predictive forecasting.  First, is a need to understand today’s market better to ensure the right customers are being reached using the right channels. Secondly, tools and methods are needed that will for the modelling of different scenarios as accurately as possible in real-time. 
We can classify and group the required tools into three main categories:
1. Integrated business planning tools (such as demand planning, supply network planning and production planning and scheduling) that simulate, plan or optimise using historical or sensed data streams to plan supply chain networks using various forecasting models. These tools also embed sales and operational planning (S&OP) processes, which overlay financial information like budgets and sales forecasts but also look at capacity constraints in production or storage facilities.

2. Agile data science tools that enable and support the creation of different models and scenarios based on various predictive analytics models such as classification, clustering and regression. 

3. Data science democratisation tools that support business analysts to run, compare and load different versions of forecasts and classifications. This will enable the business to be able to run its own scenarios and control the timing, rather than waiting for a new predictive model to be deployed. 
Getting both approaches right will enable users to know what raw materials and ingredients to buy, what inventory to hold, what finished products to make and how to efficiently ship products to the right customers. 
The ability to maximise the use of predictive planning will be key to success for all food manufacturers. Predictive planning is based on proven time series machine learning algorithms that can better predict cash flow, expenses, headcount needs, sales trends, and supply and demand. It enables financial analysts to plan the right business activities according to their cashflow forecast and avoid unnecessary spend. It also delivers huge added value by integrating with planning models to bring forecast and planning together so the planning process is a combination of actuals, budget, and forecast inputs.   
Reaping the rewards
Over time, those that have put an agile predictive planning approach in place are likely to realise greater efficiency and agility as smarter decisions can be made based on accurate predictions and trends. These choices are also based on real-time data collection from various external sources that can extend the traditional forecasting approach and cover probabilities, chances and risks to meet and avoid negative business outcomes.
Given the significant market fluctuations that have been seen in recent months, organisations have really benefited from using the approach to navigate these fluctuations, allowing them to react quickly, whether that means changing product lines or shifting into new markets. However, it is not just about churning the numbers, it is also about being creative and adaptive enough to see new opportunities and take advantage of them. Those that have done this most successfully over the past year are now seeing the benefits with increased demand or improved customer service.  
Running multiple predictive simulations can help to quickly evaluate potential impacts of different lockdown policies, which could potentially better inform critical decision-making for businesses. Ultimately, planning and modelling can plug the gaps for manufacturers and supply chains in a world where historical forecasting is far less reliable than it once was.

Richard Seel is managing director at Delaware UK.

Contact Details and Archive...

Print this page | E-mail this page