How top-performing grocery advertisers increase awareness

By: Kavya Kilari, Analytics and Insights

Story highlights:

In a 2019 study of more than 5,600 US-based companies in Amazon’s Grocery category, we observed three advertising tactics that were used more by top-performing advertisers than other advertisers. The Grocery category includes brands selling products such as Whole Foods, coffee, cold beverages, and snacks.

Top-performing advertisers have, on average, 2.2x higher year-over-year (YoY) growth in their Amazon Brand Index (ABI) awareness measures (these provide advertisers with mid- and upper-funnel metrics quantifying the number of customers aware of a brand), and 1.9x higher YoY growth in ABI consideration (which reflects the number of customers considering purchase) compared to other advertisers. To study this impact further, we used machine learning algorithms to identify tactics that differentiated top-performing advertisers from other advertisers. This article explores those tactics and provides recommendations on how to improve them.

For more information, see the Methodology section at the end of this article.

1. Top-performing grocery advertisers leverage Streaming TV ads


Our analysis shows that Advertisers in the study, on average, increased their total net media reach (over linear TV only) by +2.2% through leveraging Streaming TV ads. Moreover, a 2019 Nielsen study revealed that approximately 39% of this incremental reach could not have been achieved through linear TV because of the increasing shift from linear TV to streaming TV.


more likely to run always-on campaign


increase in net reach


of reach attribuited to Fire TV launch


Advertisers should not only consider adding Streaming TV ads to their media plan, but they should also consider running always-on campaigns for at least 25 weeks of the year.

2. Top-performing grocery advertisers use display and Streaming TV ads together


Our research showed that Streaming TV ads and display ads work better together. Top-performing brands that used display and Streaming TV Ads together saw a +47% increase in ad-attributed branded searches in Amazon’s store year over year. Taking a full-funnel approach through using complementary, always-on upper-funnel and lower-funnel strategies can help advertisers engage with customers across channels, wherever they are in their journey.

Note: While branded searches do not always result in sales, they can indicate an increase in brand consideration, which is a crucial step in the consumer journey.


Consider increasing support by running Streaming TV ads on top of display ads. Advertisers should also make use of Amazon Advertising’s tools such as:

3. Top-performing grocery advertisers use audience segments Amazon audiences more


Top-performing advertisers saw a +44% increase in consideration when reaching audiences based on behavioral signals (e.g., lifestyle audience segment) compared to when they used only demographic audience segments. Top-performing advertisers also delivered +4.5% more impressions than other advertisers.


consideration increase


impressions delivered


By using a combination of Amazon in-market and lifestyle audiences, advertisers can tailor their campaign approach. For example, when shoppers purchase in the Exercise and Fitness category, they might be doing so to help them achieve new fitness goals, so advertisers should consider using this when developing their messaging. Similarly, health-conscious shoppers might also be looking for protein powders and other products to help them supplement their diets in connection with these goals. Amazon can help advertisers reach audiences engaged in shopping activities that indicate they may have recently purchased a fitness-related product. Advertisers with access to Amazon’s Amplifier reports and Audience Insights should consider using both resources to better monitor and optimize their in-market and lifestyle audiences.


In this study, we analyzed over 5,600 brands in the Grocery category in the United States in 2020. The Grocery Category includes brands selling products such as Whole Foods, coffee, cold beverages, and snacks.

We created a composite success score of awareness Amazon Brand Index (ABI) year-over-year growth and consideration Amazon Brand Index (ABI) year-over-year growth, and then identified the top advertising and retail strategies to help increase their composite score with machine learning algorithms.

We first used a supervised model to identify a list of 20 attributes that help improve the composite score among 40+ media and retail attributes. We then uses this list of attributes to perform cluster analysis among brands, so that brands in the same cluster are similar in ad and retail attributes, while brands in different clusters are different in terms of their ad and retail attributes. These attributes include product usage such as Streaming TV ads, video ads, and Sponsored Products.

The machine learning algorithm returns clusters. We rank these clusters by the success metrics, compare the most and least performant clusters, compare their differences and conclude the key attributes that differentiate their performance on awareness and consideration ABI growth.

How are advertisers distributed across the clusters?

We used machine learning algorithms to automatically classify advertisers into clusters based on their advertising and retail attributes.

Cluster 1

Cluster 2

Cluster 3

Cluster 4

ABI awarness YOY growth. Cluster 1: 2.2; Cluster 2: 2.0; Cluster 3: 1.8; Cluster 4:1.0

ABI awarness YOY growth

ABI consideration YOY growth. Cluster 1: 2.5; Cluster 2: 1.7; Cluster 3: 1.6; Cluster 4:1.0

ABI consideration YOY growth

Cluster 1 had higher year-over-year growth in both awarness ABI (2.2x) and cnsideration ABI (1.9x) than Cluster 4. Though Clusters 1 and 2 had similar year-over-year growth in awarness ABI, Cluster 1 outperformed Cluster 2 on consideration ABI year-over-year growth (1.87x vs 1.25x, respectively)

How does clustering work?

We created a binary composite score based on DPVR, and then applied an XGBoost classifier to identify which features and which weights best predict these labels. In doing so, we considered advertising or retail actions as features such as ad product usage intensity and mix, timing of advertising support, tactics of targeting, creatives and placements, customer review counts and ratings, percentage of products with quality product pages, and the types of products promoted in ads.

Using the identified features and weights mentioned above, we then applied a k-medoid clustering algorithm to classify advertisers into clusters. Note that we classified advertisers by their actions rather than by the components of their composite score. Next, we ranked the final clusters by their composite scores from high to low. Cluster 1 is the most successful cluster with the highest composite score, and Cluster 4 is the least successful and has the lowest composite score.