Accurate forecasts are the panacea for food manufacturers, Is artificial intelligence getting us close to a solution, are we near to finding the forecasting Holy Grail?
I was at a director’s dinner recently, discussing the use of AI in today’s food manufacturing companies, when I suggested that artificial intelligence would provide the ultimate solution to forecasting and planning. I supported my announcement by describing the work being done by IFS Labs where the forecast for introducing a new product was created using automatic clustering of similar products combined with a neural network trained on historic product introduction data. The results had been very interesting with a prediction asymptotic to the actual demand compared to the seemingly random forecast based on history and ‘experience’.
The room went unerringly silent and was followed by a torrent of questions, there was a true air of excitement much like when someone knew where the holy grail was hidden. So why is forecasting such a big deal and how could AI be the solution?
Having a 100 percent accurate forecast is the holy grail in manufacturing, to know that you are making exactly what your customers want, when they want them would be a planning nirvana. The effect on a business would be dramatic, no waste, reduced cost, improved margins, increased efficiency, better customer satisfaction.. the list goes on.
But achieving this holy grail is extremely difficult because the influences on a forecast are numerous and complex. Many companies can simply plug the gaps in their forecast inaccuracy by holding stock or introducing techniques like DDMRP to hold strategic buffers. But for many manufacturers this is not possible as these solutions don’t work with short shelf life, where warehouse space is restricted, or when cash to tie up in stock is too big a risk. These restrictions coupled with an ever-increasing desire for new product ranges, in ever more variable packing formats is making the ‘hold more stock’ not an option. So, manufacturers are looking back to the source of their problems – and it all begins with the forecast.
Why is forecasting so difficult? Process manufacturing companies have been doing it for years, whether it’s a third-party forecasting solution and emended ERP capability or a home-grown spreadsheet they still don’t get it right. The main reason is that it is extremely complicated. For example, let’s look at some of the external influences that make achieving an accurate forecast so difficult.
Demand from retailers is never consistent, promotions cause huge fluctuations in demand patterns and new product introductions abound, making the use of historical based forecasting “an informed guess” at best. Retailers change their minds at the last minute, spikes or dips in depot or store consumption are reflected quickly back on the supplier with little warning.
Material supply is becoming less stable with political and trade disputes forcing manufacturers to try different sources and qualities to ensure that shortages are minimized
“Will it be sunny tomorrow?” For fresh and chilled producers the answer to this simple question could mean massive changes in customer buying patterns, statistically we simply don’t buy salads and fruit on cold wet days, and these swings can be significant sometimes well over 50 percent.
Regulation is less of an immediate effect but nonetheless can impact a company forecast, like the announcement of a ban on single use plastics. Which results in new products, different shelf life characteristics and changes in behavior of consumers.
Media is probably the biggest disrupter in demand, who would have guessed that one tv chef could disrupt a whole supply chain, but the “Delia Effect” is well known. Delia Smith a well-known UK chef has three times caused mayhem in the supply chain.
- Delia’s “Back to Basics” a cookery program, caused a surge in egg buying that emptied shelves and subsequently resulted in a sustained growth of over 10 percent
- One retailer saw sales of prunes rise by 30 percent after copycat cooks tried to recreate her Prune and Date cake
- A three-minute TV commercial was all it took to inspire home cooks to attempt to replicate Delia Smith’s latest recipe, a rhubarb and ginger brûlée. Following the commercial customers snapped up 14 weeks of the produce in just four days.
Consumers also have a significant effect on the forecast, where they change demand patterns caused by when and where we take our holidays, fads and trends driven by social media, or even by creating spikes in demand as we attend key events like festivals in the calendar. I once knew a sandwich manufacturer who serviced Motorway service stations, they changed the quantities going into the Northbound and Southbound carriageway stations depending on whether the northern football teams were playing at home or away against the southern clubs.
Artificial Intelligence – the knight in shining armor
The reason why AI has the power to help us in solving the forecast conundrum is because it has two very important capabilities;
- AI can process vast amounts of data faster than humans
- AI can utilize and assimilate multiple data sources like taking twitter feeds on food blogs, weather predictions, population movements, epos data direct from the till, population demographics and convert them from vast data lakes into statistical values and trends that a forecaster can act on.
Yes, “that a forecaster can act on.” I am not convinced that in the near future we will simply take a forecast generated by artificial intelligence and use it to run our business on a daily basis without some form of sanity check. I have many reasons for saying this, which I will cover in follow on blog. But for now, I can see a path to the holy grail and it will be interesting to see who discovers it first!
If you enjoyed this blog then please look out for others about the team at SPS in IFSBlogs
I welcome comments on this or any other topic concerning process manufacturing.
Connect, discuss, and explore using any of the following means:
- Twitter: @ElkinsColin
- Email: firstname.lastname@example.org
- Blog: http://blog.ifs.com/author/colin-elkins
- LinkedIn: https://www.linkedin.com/in/colinelkins