There is a saying that when developing a new food product, one wants the process to be good, fast, and cheap. The purpose of that is to:
- ensure a good product — that is safe and regulatory compliant, has sufficient shelf-life, and fully meets all flavor and sensory expectations of the intended target audience.
- get to market as fast as possible.
- spend the least amount of resources possible (lab time, technical testing, and support).
The dilemma is that it is only possible to achieve two of these goals:
- Good and fast will not be cheap.
- Good and cheap will not be fast.
- Fast and cheap will not yield a good product.
However, the use of artificial intelligence (AI) promises to play a groundbreaking role in the product development process to better optimize these three outcomes simultaneously.
More to the point, the literature indicates that an emerging trend in product development is for innovative food companies to harness the power of AI for this purpose. Companies strive to continuously improve their product offering with line extensions, product improvements, cost reductions, and new products which meet consumer taste preferences better and can be launched faster. In addition, consumers are desiring natural, artificial, and additive free formulations which further raise the bar on the need for an effective and efficient product development process to get products to market.
Food companies need to constantly revamp their products to cater to shifting consumer preferences. Increasingly, these shifts in consumer preferences are occurring more rapidly than the typical product development cycle. This is where AI comes into the picture. Emerging AI applications provide a unique opportunity to meet this challenge and get products to the market faster without compromising “fit for use.”
Many popular food brands have been marketed for many decades and many as far back as the 1800s. Product development files are generally not purged, including those of the numerous ingredients and trial formulations of products which never made it to the market. Consequently, there is a wealth of information in company archives on past product development activities.
This can include extensive product formula libraries, product attributes, consumer testing results, and other data points. These sources of often-unstructured data can be inputs for AI analytical tools to mine, extract, and connect relevant information in new and innovative ways and to yield insights and combinations that result in better targeting of the wants and needs of customers.
Traditional product development is essentially a trial-and-error process. A product developer, working from a brief generated by marketing, accesses the company’s ingredient library. Then, with some knowledge of past formulations, develops a bench-top prototype. In an organization with multiple product developers, each usually has their own preferences and biases regarding their “go to” suppliers and ingredients.
Product developers can be creatures of habit. There can be a proclivity to return to what works for them and to access the same base formulas and limited portfolio of ingredients they are comfortable using. This is particularly safe from a reliability standpoint when working against the clock to get a prototype out fast.
But over time these personal biases can result in sub-optimization of a prototype formulation simply because of their failure to consider the full range of what is generally a treasure trove of available raw material inputs. Such a blinkered approach will obviously not yield the best solution.
AI can be harnessed to enable the product developers to apply analytics to scan the complete set of available inputs, including a broad base of past and present formulas, ingredients, as well as other data.
The tool can be applied to eliminate bias; identify combinations a human developer could not necessarily conjure up; and, in doing so, provide a greater likelihood of bringing forth the optimal solution.
Once a bench-top prototype is developed, it is subjected to panel and/or consumer evaluations to determine if it meets the desired flavor profile criteria which is the next logical step in most stage-gate product development processes.
More times than not, at this point, the prototype is often rejected and sent back to the lab for another round of iterations to refine the product. Quite often, it can take multiple iterations of returning to the bench for formulation refinements before there is a settlement on a final formula that sufficiently hits the mark. But only at that point can product development proceed to the next step — which may involve consumer or customer evaluation. Needless to say, it burns considerable time, money, and resources to home in on the desired product.
The promise of AI in the product development process is the ability to access the company’s legacy data more efficiently, effectively, and rapidly to produce a product which is optimized for broader and more sustainable consumer appeal. The result is greater speed to market, lower product development cost, and products which expand the realm of “what is possible but not obvious.”
Explore the July August 2020 Issue
Check out more from this issue and find your next story to read.
Latest from Quality Assurance & Food Safety
- FDA and EPA Announce First Registered Pre-Harvest Agricultural Water Treatment
- USDA’s Agricultural Research Technology Center Breaks Ground in California
- Submissions Open for Fourth Annual Seeding the Future Global Food System Challenge
- PPM Technologies Introduces FlavorWright SmartSpray Food Coating System
- Mettler Toledo Unveils New X52 X-Ray Solution
- FDA Issues Final Compliance Policy Guide for Scombrotoxin (Histamine)-Forming Fish and Fishery Products
- World Food Prize Foundation Announces $50,000 Innovate for Impact Challenge
- Ron Simon & Associates Retained by 33 Victims in McDonald's E. coli Outbreak