Label Insight’s Trending Attributes feature maps search trends to products to help brands optimize the digital shelf. Using Label Insight’s industry-leading product metadata, the Trending Attributes feature knows what attributes every product qualifies for, even if the brand is not explicitly making a claim about that attribute on-package. And by layering in consumer search data from leading e-commerce retailers Amazon, Walmart, Target, Instacart, Shipt and Kroger, Trending Attributes knows which attributes are most searched, and maps those search trends to products for actionable takeaways like content optimization, improving organic search ranking, or signal identification for whitespace innovation trends.
The Product Metadata
Label Insight works on behalf of Retailers with CPG Manufacturers to create and catalog granular product data. Every piece of information is digitized from the package, and from that base layer of raw claims, ingredients, nutrients, and more, Label Insight catalogs up to 24,000 attributes per product through advanced ingredient properties, claims taxonomies, data science, and a team of registered dieticians, nutritionists and chemists with intimate knowledge of government regulations & criteria for making certain claims on-package.
Label Insight offers 2 types of product attributes: On-Package and Derived. On-Package attributes tag products based on the presence of a claim or certification on the package label. Derived attributes tag products based on whether an attribute is inherently true for a product based on Label Insight’s deep analysis of the label, including insight into nutrients, ingredients, and regulated allergen & warning statements.
The Search Data
Label insight works with a search panel partner with over 3 Million US consumers to gain insight into the direct, on-site searches consumers are typing into the search bar on sites like Amazon, Walmart, Target, Instacart, Shipt & Kroger.
Prior to Label Insight receiving search terms from our partner, the data undergoes a lengthy process:
- Collected via statistically representative datasets from over 3 Million US consumers.
- Synthesized leveraging a sophisticated algorithmic process to clean, match, synthesize, process, and blend inputs for data modeling.
- Modeled using a normalized dataset, and run through advanced machine learning calibration & predictive models to provide an accurate and consistent view of the digital world over time.
This machine learning for synthesizing & modeling search data includes, but is not limited to:
- Cleaning data to remove any Personally Identifiable Information at the source and to format data inputs
- Classification of data inputs for categorizing and synthesis
- Synthesizing billions of data inputs for advanced, predictive modeling
- Training machine learning models and refining for noise and bias reduction
- Blending models for weighting & projection, scientific calibration, and delivery
*Special Disclaimer on Long Tail Search Volume
Label Insight receives unique insight into unmasked volumes for long tail search terms that include attribute & category-specific keywords, such as “paraben & gluten free shampoo & conditioner”. Many of these long-tail search terms fall below the 5,000 searches threshold, which is generally considered to be statistically significant for digital market measurement purposes, and is intended for directional purposes only.
The Trending Attributes Advantage
Label Insight’s advanced machine learning has been trained on over 47 Million unique on-site search terms from Amazon, Walmart, Target, Instacart, Shipt & Kroger. These incoming search terms undergo a lengthy data science & QA process to remove noise and focus in on attribute specific searches:
- Cleaning incoming search terms through a complex series of filters to bring in only terms relevant to a product type Label Insight supports (e.g. Food & Beverage, Personal Care or Pet Food).
- Classification of cleaned terms through a recognition process including data science model confidence in a term being relevant to an “E-Commerce Category” (e.g. Bath Preparation & Body Wash). Terms must be explicitly recognized as belonging to an “E-Commerce Category” either via highly confident data science models or a review & model retraining process done by Label Insight’s team of subject matter experts.
- Keyword Mapping & Aggregation of classified terms through a keyword mapping process of search terms to attributes. An Attribute taxonomy considers both “include” and “exclude” keywords, and is maintained by subject matter experts at Label Insight to define keywords relevant to an Attribute (e.g. “gluten free”, “gluten-free” and “gf” may all be included keywords for the “Gluten Free” attribute, and “no calcium” may be excluded from the “Calcium” attribute). Terms must be explicitly included or excluded as belonging to an Attribute, either via highly confident data science models or a review & model retraining process done by Label Insight’s team of subject matter experts.
- Structured Analytics Reports of search volume for aggregated attribute keywords across a product type like Food and Beverage, or specific to an e-commerce category like Bath Preparations & Body Wash allow unique insight into which attributes are most important and which are growing/declining in popularity. These reports are updated monthly with new search data, and are made available within 2-3 weeks after the end of the month.
- Product Attribution allows Label Insight to map search trends back to products. This provides unique insight into which attribute trends are relevant to a brand’s portfolio of products, which of these trends the brand is already taking advantage of by making an on-pack claim, and which attributes represent unclaimed opportunities where the brand could be making a claim on pack, but is not.
- Workflow & Integrations in Label Insight's Attribute Management Platform allow regulatory managers to enable or disable attributes across an organization, or e-commerce managers to add attributes or keywords to their item setup templates for syndication.
Article is closed for comments.