Why you should consider using negative targeting in your next campaign
By: Zhixun Wang, Senior Data Scientist, and Ashton Brown, Technical Writer
In this study of over 3,500 brands in the Home Environment category in Amazon’s store we compare advertising strategies of top- and lower-performing advertisers. We then use this comparison to derive actionable insights advertisers can use to help improve new-to-brand customer growth rate and detail page view growth rate year-over-year.
Story highlights:
In this study, we analyzed over 3,500 brands in the Home Environment category in 2018, 2019, and 2020. The Home Environment category includes brands selling products such as air purifiers, humidifiers, and heaters. To perform our analysis, we grouped Home Environment brands into five clusters, with Cluster 1 being the most successful in terms of detail page view year-over-year growth rate (DPVGR), return on ad spend (ROAS), and spend per customer (CS), and Cluster 5 being the least successful; we define CS as the average sales amount per customer.
Our analysis finds that top-performing Home Environment advertisers (Cluster 1) had a 1.1x higher DPVGR, 50% higher ROAS, and a 16.5x higher CS in comparison to lower-performing advertisers (Cluster 5).
Top-performers
Higher DPVGR
Higher ROAS
Higher CS
To provide advertisers with actionable insights, we used machine learning to analyze 50+ advertising attributes that contribute more or less to DPVGR/ROAS/CS. We then identified which attributes have the largest positive impact on the year-over-year growth rate of detail page views, return on ad spend, and cost per customer. This article provides insights and best practices on the key attributes or strategies by quantifying the degree to which top-performing Home Environment brands (Cluster 1) and lower-performing Home Environment advertisers (Cluster 5) have adopted each key attribute or strategy.
For more on how we performed this study, see the Methodology section at the end of this article.
Top-performing Home Environment advertisers used negative keywords and negative ASINs 4x more than lower performers
Relevance in shopping results may lead to higher consumer engagement. On average, 12% of impressions were delivered in campaigns with negative keywords or ASINs tactics in the cluster with the highest success metrics, but it was 3% among those with the lowest score.
Percentage of impressions delivered with negative keywords/ASINs
Top-performers
Lower-performers
Examples of using negative keyword or ASIN tactics
To showcase how negative keyword or ASIN tactics are used, we will look at two scenarios:
- Scenario 1 - A company promoting home air filters can include negative keywords such as “automotive” in their campaigns. This will prevent ads from competing with car filters, which means that campaigns dedicated to home air filters will show up in response to more relevant shopping queries.
- Scenario 2 - A company that wants to advertise their tower fan line only can use “desk” or “desktop” as negative keywords to avoid showing ads to shoppers looking for a fan for different usage.
Learnings and best practices when using negative keyword or ASIN tactics
To showcase how negative keyword or ASIN tactics are used, we will look at two scenarios:
- Use metrics for selecting negative keywords: Lower click-through rates and lower conversion rates are good indicators of keywords that are underperforming and could be the source of candidates for negative keywords.
- Test and learn: Test negative keywords that work in theory, learn, and frequently optimize with the ones that are performing.
Top-performing Home Environment advertisers were 4.5x more likely to use geographic audiences than lower performers
Our analysis found that top-performing Home Environment advertisers used geographic audiences to deliver 9% of their total ads, while lower performers only delivered 2% of their ads through geographic audiences. This approach can help campaigns reach more new audiences in specific regions. It can also help campaigns reach audiences regions with different needs for Home Environment products. For example, advertisers may want to show fan and dehumidifier ads in hot and humid areas and show heater and humidifier ads only in cold areas during winter months.
Percentage of ads delivered with geographic audience segments
Top-performers
Lower-performers
Learnings and best practices when using geographic audiences
- Create custom audiences: Consider leveraging Amazon Ads tools to create custom audience segments based on geographic signals that align with campaign objectives.
- Seasonal effects: Consider using appropriate seasonal geographic tactics for different products based on different climates across geographic regions.
Conclusion
Our analysis found two key ways advertisers can positively impact their DPVGR/ROAS/CS. First, advertisers can use negative keywords or ASIN tactics to ensure that their ads are reaching their desired audiences according to product type. Second, advertisers can use geographic audiences to ensure that their ads and products are being seen by customers - according to where they are. Not only can the two methods be helpful in improving DPVGR/ROAS/CS individually, but they can also benefit each other when working together.
Methodology
We first used a supervised model to identify a list of attributes that help improve the composite score among 40+ media and retail advertising attributes. Specifically, we followed a five-step process to create a success metrics of detail page views year-over-year growth rate, return on ad spend, and average spend per customer, and then identified the top advertising and retail strategies to help increase the success metrics with machine learning algorithms.
- Select brands: 3,500+ brands in the Home Environment category in 2018, 2019, and 2020.
- Create success metric: Calculated based on year-over-year growth of detail page views, return on ad spend, and average spend per customer.
- Identify effective ad or retail actions: Identified top actions to help increase composite score (actions that lead to higher year-over-year growth of DPVGR/ROAS/CS).
- Group brands: Brands grouped by composite score (DPVGR/ROAS/CS) into five clusters ranked from highest to lowest performing.
- Compare brand groups: Identified what strategies they used and quantify by how much top-performing brands (Cluster 1) increased year-over-year growth of detail page views, return on ad spend, and spend per customer compared to strategies that lower-performing brands (Cluster 5) use.