Background
Over many decades the FMCG industry has invested heavily in tech. During the last 50 years, from EPOS to retail media, many suppliers and supermarkets have invested in tech to achieve better outcomes. Numerous results highlight that tech investments don’t always work in FMCG (and probably other industries).
An obvious example of tech failing is that shoppers are buying more private label products at hard discounters. These SKUs are usually copies of branded products (no innovation) and are not supported by tech (e.g. viral social media posts). Hard discounters generally have simple operations to minimise CODB (cost of doing business) and are not tech driven (e.g. no loyalty cards, click and collect, e-commerce). The continual sales growth of these simple products in stores with a simple offer highlight what many shoppers now want – value, not necessarily tech.
Why AI Will Struggle to Deliver in FMCG (Despite the Hype)
The tech industry suggests that AI could develop into a system that manages complex systems – such as FMCG supply chains. This outcome would be achieved by analysing large amounts of data quickly and then implementing changes to deliver an optimal outcome.
Many suggest AI will dramatically change FMCG (and other industries). Firstly, the FMCG industry (suppliers, supermarkets) has already automated many functions from demand planning to customers self-scan in store. Automation, via tech, is not a new idea in FMCG. Secondly, the shopping experience is already tech driven. Many shoppers discover products online, order online, get home deliveries etc. To argue AI (or any other new tech) will dramatically change the industry is debateable. What will occur is that tech will continue to evolve and so will the FMCG industry and shopper behaviours.
The fundamental reason AI will struggle in FMCG is because the industry is supplying a physical product via a complex and large supply chain. Other industries, e.g. music, have transitioned from physical (e.g. CDs) to digital (e.g. streaming services). FMCG is built around physical supply chains and stores (including dark stores for e-commerce) serving shoppers. These supply chains are complex, challenging (e.g. COVID, war in Iran) and driven by shopper demands. Shopper demands are not logical (e.g. toilet rolls during COVID, petrol now) but are driven by emotional and physical demands. Also to implement AI recommendations requires collaboration between suppliers and supermarkets.
Shoppers drive sales / supply chain
As explained in my book, Category Management is a way of thinking , shoppers have increasing power in FMCG. This has been occurring for decades, and tech is continuing to increase shopper power (e.g. online reviews). AI will struggle to understand why people make decisions.
For example, due to the war in Iran petrol prices have increased. A logical response would be shoppers buy less petrol, but the opposite, panic buying has occurred. During COVID shoppers panic bought toilet rolls and for a brief period shoppers (particularly teenagers) wanted Prime Hydration drink. These examples highlight a constant challenge in FMCG – human behaviour. The industry (and the tech it uses) can never accurately predict shopper behaviour. After the fact we realise what occurred, why etc but no predictive model accurately predicts shopper behaviour.
AI models rely on historical data (algorithms) to predict future shopper behaviour. Numerous factors affect future shopper behaviour (from cultural nuances to changes in in-store execution) and no model will ever accurately capture, analyse and importantly give sufficient notice to suppliers and supermarkets to change their plans to meet the change in shopper demand. For example, AI is available today, but petrol stations are still running out of fuel.
Data quality / actionable insights
AI models rely on data for their output. Much of the data required for fact-based decision making in FMCG is controlled by suppliers and supermarkets. Will Coca Cola and Pepsi share their data / insights with AI so the drinks industry can make better informed decisions? Will Coles and Woolworths share all their data (for free), down to store / SKU level for all other supermarkets to see? The reality is that AI will work with limited data in FMCG. The publicly available data is fragmented across different suppliers and supermarkets and of little commercial value.
The reason why suppliers and supermarkets collect and analyse large quantities of data is to better serve shoppers. The commercial value of this data (insights) is significant, and they won’t share it publicly. Suppliers and supermarkets may use AI (and other tech) to better manage and analyse their data internally. All practitioners understand the limitations of the data available. Only basic data is captured. For example, in store execution (off-locations, planograms) is not recorded, just unit sales, price etc. So was the Coca Cola off location promo successful due to price, in store (e.g. physically large displays), external (e.g. hot weather) or Coca Cola trending on social media? The reality is nobody knows the exact answer. The scan highlights basic facts like units sold. Loyalty card / panel data can add more depth, e.g. penetration. The challenge is quantifying psychological factors like a war in Iran changing human behaviour.
AI models in FMCG will have limited, incomplete data to analyse. This means AI will struggle to deliver actionable insights. Human interpretation of the data will continue (😊) due to the simple fact that humans have a better understanding of human behaviour and suppliers / supermarkets will keep the data in house.
FMCG complex large supply chains
There are numerous examples of major disruption to FMCG supply chains (e.g. COVID, war in Iran). These examples highlight just how large and complex FMCG supply chains are. For example, a war in Iran can increase the price of tomatoes in Australia. In addition to these major disruptions there is constant change in FMCG supply chains. Examples of constant changes include changes in fresh availability (quantity, quality, price), new product launches, price promotions (including long-term EDLP), packaging changes, in store execution including off location displays, retail media and planograms.
The constant change in FMCG supply chains (from minor to major) make AI models application limited. For example, what historic data can be modelled to understand shopper demand for petrol due to the war in Iran? At a category level how will be a packaging change (new design) for brand A effect sales of brand B?
Constant change, in large complex supply chains, will limit the quality of AI generated recommendations for FMCG too. For example, AI may suggest to source more fuel for petrol stations now but how can that physically be achieved?
IMHO (in my humble opinion) the numerous issues that Amazon has encountered running supermarkets highlight the complexity of the FMCG supply chain. Amazon is definitely a global leader in e-commerce (with deep pockets) but has struggled with grocery. The closures of Amazon Go, Amazon Fresh stores highlight the challenge. Simply put, the retailer with (arguably) the best tech in the world has struggled to manage a FMCG supply chain.
4 . Supplier / supermarket collaboration
FMCG products are sold via numerous channels (D2C, e-commerce, foodservice, HORECA, QSR, petrol and convenience, supermarkets) to shoppers / consumers with different demands. The success of these complex, large supply chains is often determined by the relationships between suppliers and their customers (including supermarkets). As outlined in my blog, There is no perfect category relationship , factors such as trust are important in these relationships.
Supplier / supermarket collaboration is key to implementing change in FMCG. Recommendations that are driven by AI, but not supported by practitioners, will not be successfully implemented. Simply put when suppliers and supermarkets collaborate (agree and implement a plan) the overall outcome can be maximised. If suppliers and/or supermarkets do not agree with AI recommendations’ then they won’t get implemented.
AI Cost Benefit to shoppers
Hard discounters, such as Aldi, have deliberately minimised tech (and associated costs) in their supply chain to offer shoppers value. Shopper based tech, e.g. home delivery and loyalty cards, create complexities in the supply chain and increases CODB (cost of doing business). Aldi shoppers simply don’t want to pay for this tech.
AI may be able to generate more insights, faster etc but what will the overall cost to the supply chain be? With shoppers increasingly demanding value, due to COL (cost of living) pressures, what is the cost benefit to shoppers of implementing AI?
Despite the hype, the benefits to shoppers of new AI models is still an unknown. The success of simple operations, such as Aldi, highlight the risk of suppliers and supermarkets investing in unproven tech that shoppers will not want to pay for.
Summation
The tech industry suggests that AI could develop into a system that can manage FMCG supply chains. This is debateable due to:
Human behaviour constant changing
Lack of data (quantity and quality) for models to interpret
Difficulty implementing changes in a complex and large supply chain of physical products
Challenges in having all supply chain parties (e.g. suppliers, supermarkets) aligned to the same plan
Shoppers be willing to pay for AI
The FMCG industry will continue to invest in tech to manage their operations and better serve shoppers. AI will be part of the FMCG industry but it is unlikely to evolve into a system that could replace the current system.