External Methodology

SPOONSHOT METHODOLOGY

OVERVIEW

At Spoonshot, we are building a next-generation data tool that can transform food innovation by combining artificial intelligence and fundamental sciences to transform the new product development journey. We have developed technology that processes large volumes of unstructured data, weaving together seemingly disparate information to uncover hidden relationships and opportunities.

Data is the new oil. Having access to differentiated data sets from your competitors is a serious advantage for genuine product innovation. Food is very local in nature, however, an idea or inspiration for new product development can come from any part of the world as food trends travel across regional boundaries.

Here are a few guiding principles in our data strategy:

Diversified sources: Data is being generated by everyone and everyone is hungry for data. It is important to gain a competitive advantage and so have as many different types of data in your grasp as possible e.g. industry events, blogs, patents, media, and user-generated content, etc.

Authenticity: Every data source goes through stringent checks, both manual and automated, where we examine factors like quality and completeness of the content, rank, brand, age, social proof, etc.

Go local globally: We look at long-tail of data sources where each of them is a local source to a given region. This helps us spot potential trends early on and report this to you based on your information needs.

Unification of all data silos using domain knowledge: The key to derive an insight is based on the ability to connect seemingly disparate information dots. Our unification agent is food science, as detailed above. It not only allows us to identify opportunities early, it also means that we can give every different type of food business the necessary context of an emerging trend, and how they should act.

More than 80% of the internet’s data is unstructured in nature, and it comes in various forms from a multitude of sources. Also, the rate of the growth of data is only increasing. Did you know that 90% of the world’s data was created in the last 2 years, with food being the most searched and talked about topic ?!

TAKING A DEEP DIVE LOOK AT DATA ACQUISITION

Our technology vision is to build an explainable machine intelligence that can replicate human cognition in food.

The key highlights of this approach include:

Connecting disparate dots: We keenly observe the current world of food and beverage and then analyze co-relations to it - such as consumer food interests, the impact of weather or the regionality of a particular cuisine, creating a pattern that provides more actionable insights.

Powerful combination of domain expertise and tech: Having domain knowledge is not just enough. Our technology combines both domain knowledge with structured data in such a way that we are able to derive sense out of it.

Confluence of Food Science & AI: We use AI techniques, like Natural Language Processing and Computer Vision technologies, to build structured information from unstructured data.

We do this by leveraging food science principles to establish relationships between these information dots. Depending on the application or insight being delivered, the appropriate datasets and connection types are used.

We collect information relating to the physical and chemical properties of ingredients and understand how ingredient interactions impact a final recipe, to build insights that are valid and relevant to you. Currently, our database contains volatile compounds for each ingredient, their sensory flavor profile, and the nutritional breakdown of ingredients. We use publicly available data e.g. web, academic articles, and patents, to get this information.

THE NEXT TREND

We monitor more than 2,900+ diverse long-tail sources and dive into the archive of product data, brand perception metrics, scientific journals, niche food communities, social commerce platforms, and consumer conversations. Based on this data, we separate isolated patterns from underlying cultural movements and analyze the signals. We constantly track these signals and identify how they’ve evolved every month.

Because these long-tail data sources are all vetted, you can be assured of their authenticity. The inference of trends is determined by monitoring the changes aggregated over these long-tail sources.

We try to answer the ‘why’ behind any trend that we find. Any long-lasting trend is usually driven by a bigger macro or megatrend. For example, the growth of popular plant-based products is because of sustainability concerns and the rise in chronic diseases. We use our proprietary expert knowledge base to draw the causal relations in combination with various data sources (articles, blogs, news, scientific publications).

HOW DO WE FORECAST A TREND INTO THE FUTURE (6 MONTHS)?

We have built a comprehensive dataset that comprises more than 10 years of all digital content in food. Also, this dataset grows with time as new content gets generated. This dynamic dataset serves as training data to build models to predict the future trend using historical data.

The example below shows the historical trend line for Kokumi and a forecast on its growth.

If you have any further questions about our methodology, please email .