Avoiding planning pandemonium in a pandemic: How can you forecast in the new normal?

By David Hawkings, Senior Vice President, Europe, at Antuit.

During the pandemic, the foundations on which retailers, brands and Consumer Products companies rely to determine trade volumes, pricing and promotions have been shaken and, in some cases, broken.

Legacy demand forecasting is no longer a dependable indicator because it relies too heavily on historical performance, which has been shown to be unreliable for some years, now has failed completely in the face of Covid-19. And new consumer behaviours have emerged, and are continuing to emerge, making it impossible to create a forecast that is based on demand from a time that has now passed. Some companies already recognised the need to start researching what consumers would want before making forecasting decisions, but most lack the data, the technology and the supporting processes to put these aspirations into action.

Human instincts are no longer enough to anticipate consumer demand

Brands and Consumer Products companies do rely on data from third parties, such as retail point of sale transactions, to improve their forecasting. Some will add limited external factors to make these plans more realistic, weather being the most obvious. Unfortunately, the conclusions that are drawn from this data are often not scientific, but based on instinct. Inputs from market research are used to support decisions because everyone ‘has a good feeling about this one.’

The problem is then compounded through trade promotion planning, which is driven by the same data inputs, but is performing very unreliably in the face of customer behaviour during the pandemic. Margin can be eroded when optimal pricing is missed, or sales volume expectations are simply not achieved. There can also be longer term damage for brands that are now perceived by the consumer as less valuable to them because their promotional pricing doesn’t align with their expectations.

Demand forecasting looks sophisticated enough because it does take into account a number of data points - product lifecycle, seasonality, trends, market entry, holidays, promotions and price changes - but they are all based on historical performance. The crucial element that is missing is consumer insight, the only indicator that will enable businesses to build what is called “consumption sensing”.

However, historical demand indicators should not be abandoned completely, but rather complemented by new consumers-based indicators. Employment and economic indicators, competition, loyalty programmes, social sentiment, ecommerce and performance of other channels, such as marketplaces, all now have a role to play in new forecasting.

Forecasting consumer demand will require Artificial Intelligence

If this sounds complex and it sounds like too many inputs, for a traditional planning process based on spreadsheets and instincts, it absolutely is.  This is why artificial intelligence (AI) is required to process all these inputs to create plans that can sense future consumption. Brands and Consumer Products companies are making manual interventions during the current crisis, but these are not sustainable.
 
This is not just about a single, overarching plan, but about hyper-localised models that AI has the power to create with speed and at scale which is critical as brands confront the challenge of traditional channels diminishing in importance and new ones emerging. For example, contrast the traditional supermarket with Amazon, specialist marketplaces, home delivery and social media acting as a sales channel. Consumption sensing through this many channels cannot be modelled without AI.
 
And it doesn’t stop there. The agility to act quickly and make changes must be built into these new plans. Insight can only be provided by machine learning which can react quickly to actual performance and varying inputs that have been shown to be wildly variable under Covid-19.

The goal is to have a ‘unified demand signal’. A single version of the truth that feeds four key areas which neatly integrate a number of drivers that are currently, by different teams at different times, using different data inputs:

• Consumption sensing - Growth analysis and demand driver elasticities
• Revenue growth management - Growth optimisation with price, promotion and assortment
• Demand planning - Shipment forecasting and inventory optimisation
• Sales and operations planning - Growth projections, profit and loss review and sales performance

Any suggestion that the landscape of demand will settle into an old routine within months, or even years, post Covid-19 is wishful thinking. This may work for some categories, but for the vast majority, they will have to respond to consumer behaviour that is going to be changing continuously.

Changing travel patterns, cultural fusion, new types of employment, possible redundancy in young industries, political polarisation - while we may not hope for these things, they have been emerging for a number of years and pandemics are a massive shock to how they are behaving right now. And most people now agree that some things will never return to normal. History is not dead, but it is now only one of the indicators of what consumers will do next.

Longer term, consumers will expect to influence and have closer relationships with both brands and Consumer Products companies. A survey by a tech consulting company, True Fit, showed 47% of millennials would welcome the chance to be involved in new product design. When consumers provide valuable insight into their aspirations and tastes, businesses can shorten the new product development lifecycle significantly.

Arguably, there is a fail-fast element to this approach, but it does enable brands to innovate more quickly and more often, as it searches for the next must have products.  Most importantly, anticipating and producing what consumers will expect and demand – next month or next year – is the key to success.