Digitalization might someday streamline the global sourcing of coffee and tea. Five big companies have bet that a startup’s technology will let buyers screen products without physical shipment of samples.
As the global tide of Covid-19 recedes, what remains is a world of digitalization. And it’s not just Zoom.
In the vast business of sourcing food ingredients, the pandemic exposed a major weak link in supply chain operations — the need to continually sample products like coffee, tea, cocoa, herbs and spices, and grain.
Buyers need to obtain physical samples of these goods in order to conduct chemical tests and taste tests. And the buyer often must order and screen many different samples in order to find a supply meeting very narrow, very specific characteristics. Chemical analysis is needed to confirm authenticity and purity. Then tasting is needed.
The process is complex, time-consuming, error-prone, and expensive, especially the final step of protocoled tasting by highly paid experts. The pandemic made the process even more difficult and expensive, due to bottlenecks and barriers in cargo transport and travel.
Digitalization looks like an increasingly ready remedy, particularly the new artificial intelligence and sensor technologies being developed by researchers and entrepreneurs.
One leading example is ProfilePrint, a Singapore-based startup established in 2017. In August, it won a strategic investment from agribusiness giant Cargill following six months of pilot projects that validated ProfilePrint’s technology across an array of ingredients.
The company had already attracted investments from top venture capital funds, the Singapore government, and several strategic corporate investors in the tea, coffee, agribusiness, and food ingredient sectors: Netherland-based Louis Dreyfus Company, Singapore-headquartered Olam International, Swiss coffee merchant Sucafina, and Indonesian conglomerate Sinar Mas.
ProfilePrint says that its algorithm can rapidly authenticate a product, detect adulteration, and even assess flavor using sensor technology and chemometrics, the science of using data to measure chemicals.
“ProfilePrint is on a mission to transform food ingredient supply chains globally,” said Alan Lai, founder and CEO. The company uses its patented artificial intelligence to record complex chemical parameters and sensory data, combining them into a single “fingerprint” that indexes the chemical and taste profile of a product sample.
Yet the process is about as difficult as making toast. A tea grower just needs to pop a tablespoon of leaves into the drawer of the countertop analyzer, which is about the size of a K-pod coffee brewing machine. Close the drawer and wait a few seconds for sensors to record raw test data. That information gets uploaded to the cloud and analyzed by the company’s software. The tea seller can then access the completed product profile online by smart phone or laptop and send it to a buyer anywhere in the world.
“What we saw was a lot of inefficiencies in the current supply chain,” Lai said. “Growers, whenever they produce good quality products, are not typically getting a price based on quality, but often based on quantity. This enables middlemen to up the price and earn quite good money from consumers.
“ProfilePrint is a better way to assess quality — firstly, more objectively, and secondly, more rapidly since you do not have to send it to a lab. It’s not difficult to operate. We increase efficiency without removing the value added by middlemen,” Lai explained.
The technology doesn’t make tasters redundant. A buyer, after using the digital profiles to screen out supplies that score low on the desired characteristics, still needs to conduct protocoled tasting of qualified samples. Tasters complete the process by scoring flavor, aroma, color, clarity, mouthfeel, and other organoleptic — a.k.a. sensory — properties.
Lai said that ProfilePrint is initially applying its technology for use in the supply chains of a handful of major commodities and ingredients that are prone to counterfeiting and scams: tea, coffee, cocoa, rice, other grains, spices, and traditional Chinese medicine.
Lai, an aeronautical engineer by training, spent his early career in the United Kingdom, China, and Singapore in mergers and acquisitions, private equity, and venture capital. As a daily drinker of tea, Lai recognized that repeatedly sampling tea across several time zones could be partially automated, an idea that was quickly embraced not by tea professionals but by coffee businesses, which provided $1.4 million to finance early development.
“We are democratizing expertise in the AI models, such that even a farmer who is not an expert in grading but an expert in growing tea, will now be able to assess straightaway the profile. This unbiased reference point enables the tea maker to share profiles with potential wholesalers before they ship samples for buyers to see, touch, and taste,” said Lai.
Whatever his love for tea, Lai found it easier to work with coffee in Profile Print’s early days. Coffee was less of a challenge, thanks to cupping parameters that have been well established by the Specialty Coffee Association. SCA’s methodology provided a taste profile that ProfilePrint’s technology could use as a framework.
“When we started with tea, we wanted to standardize a profile to give traders, farmers, and consumers a common reference point. Half a decade ago, maybe the industry wasn’t quite ready,” Lai said.
He expects that the tea industry will follow the coffee sector’s path in developing its own parameters, despite the many different types of tea available, and that AI can build on this system to assess samples.
“Brokers who establish, let’s say, 10 parameters that essentially define their ideal tea can use these parameters to build and train a model to identify variations. It’s customizable. Value addition comes from the expertise of the user. The more you train it, the more value you get out of the model,” he said.
When wholesalers precisely describe the profiles they are seeking, “you do not need to go through the process of sending samples, confirming arrival, then tasting different lots while asking again and again if they are satisfied before you complete the order,” Lai said.
“Traveling itself is a huge cost, whether it’s due to Covid restrictions or whether it’s due to adding carbon emissions to the environment,” Lai observed. “The world of digitization saves costs.”
Authentication
Currently, tea industry buyers use product labels to authenticate product, but labels are prone to misrepresentation and fraud. Wholesalers that fail to frequently analyze samples sometimes discover that a tea has been adulterated. Some suppliers dry mix processed tea with previously brewed “spent leaf.”
Consumers are often deceived into buying tea labeled, for example, "100% Darjeeling," when it actually includes as little as 20% genuine Darjeeling. ProfilePrint can recognize tea leaves of the same type and differentiate between different tea regions, production methods, and even growing seasons.
“The methodology allows us to authenticate the tea leaves sold by retailers or wholesalers,” Lai explained. “ProfilePrint can also trace specific chemical compounds that exist only in, let’s say, a Vietnamese city. Otherwise, you may need to go to a lab with expensive testing equipment. They take a longer time, they are destructive, and they obviously require experts to be able to do that.”
From mass to class
Digital product profiles can be used in sourcing specialty teas as well as commodity product.
Multinationals that sell mass-market blends need to sample to ensure consistency because they source many different teas in large quantities from growers at several origins. Seventy-five percent of the world’s tea is sold at public auction. In Mombasa, Kenya, for example, auctions offer some 450,000 metric tons of tea per year. Weeks before each auction there, and at similar events in Sri Lanka and across India, sellers prepare and share thousands of samples with bidders.
Brokers and corporate buyers value each lot based on quality, freshness, distinctiveness, style, and the price paid for similar teas in the previous auction. They also factor in historical averages based on sales in previous years. In addition, each lot has a unique number, a factory mark, grade, invoice number, type and quantity of packages, and net weight.
Digital profiles can help streamline and objectify much of this data to make pricing and selection easier.
Digital profiles can also help screen samples of specialty leaf in the direct-sale process, where buyers contact sellers privately, negotiating price and quantity directly. For a grower to receive full value, the sampling must be done skillfully. Buyers need to sample to ensure that a premium tea will meet the expectations of consumers.
Prospects
It’s too early to say whether ProfilePrint will succeed in its mission, or be sidelined by competitors, or discover that its technology simply doesn’t find enough customers.
But one big believer is Cargill. In a press release, Francesca Kleemans, managing director for Cargill’s cocoa and chocolate business in the Asia Pacific region, said that ProfilePrint “holds the potential to transform the global food-ingredient supply chain, strengthening the sensory innovation capabilities of our ingredient portfolio without compromising on taste and quality. This can help Cargill deliver against our high standards for food quality and enable faster and more precise product development for our customers.”