Search Term Analysis

Author: Robin Jia，Tommy Lin

Last Update：2024/09/26

1. Model Introduction

Through this model, the following business questions can be analyzed:

• What search behaviors do users exhibit when purchasing branded products, and how do ads perform across different placements?
• What other terms do users search for after searching for a specific term?

2. Model Interpretation

The model mainly consists of two modules. The first part records the traffic data of users searching for specific terms and their conversion behavior, while the second part shows the relationships between search terms, indicating which terms are frequently searched by the same user consecutively.

In the predefined model, you can select the time period fixed to a monthly data cycle. You can also choose the type of search term, either keyword or product. The default lookback period, i.e., the interval between search behavior and conversion behavior, is set to one day.

The overall section includes common filters used in search term reports, such as text search, store, ad type (Sponsored Products or Sponsored Brands), targeting, and match type (exact, broad, or automatic targeting). You can also filter by ad placement to view conversion performance across different placements.

Unlike Amazon's native search term reports, the search term report by ad placement is unique to AMC data. By analyzing the combination of search terms and placements, you can identify the most effective keyword placements and adjust bid weights accordingly to optimize ad performance.

For instance, if you want to identify the top-performing search terms for specific ad placement, you can select the placement, such as "Top of Search," and sort the data by conversion rate in descending order to find the best-performing search terms, which can guide keyword strategy optimization.

Alternatively, if you wish to analyze a specific search term's performance across different placements, use the search function to locate the term and filter by placement to compare its performance across placements. This helps adjust bidding strategies and allocate more of the budget to the best-performing placements.

A filtering feature for metric values is also available, allowing you to define multiple metrics for cross-filtering, which can help eliminate search terms with low search volume or identify those that perform well across various metrics.

The bubble chart displays the top-performing search terms comprehensively (default to top 20 by exposure-to-conversion rate and click-through rate, but adjustable), with bubble size representing search volume. You can zoom in on specific areas by dragging the axes or using the mouse wheel, making it easier to focus on key terms.

Search terms positioned closer to the top-right corner on the default metric axes indicate better performance, and using the data filter can help exclude long-tail terms with low data volume, allowing you to identify core search terms.

The lower section provides a detailed performance data table for each search term, with options for summarized and detailed modes. The grouped mode shows performance for each search term, while the ungrouped mode displays performance under different targeting. The predefined model allows a comparison with data from the previous month, and you can hide the bubble chart, expand the table to full screen, download the data, or customize columns.

Through custom columns, you can select more conversion data provided by AMC, much of which is unique to AMC. This data is not only based on user-level granular data but also covers additional conversion metrics such as ATC (Add to Cart), NTB (New to Brand), DPV (Detail Page Views), etc., providing more comprehensive consumer insights.

You can also click on the button right to a search term to view a detailed data pop-up showing weekly performance trends within the selected data period, supporting data download.

The related search section includes filters for searching associated search terms and selecting one or more search terms to view related terms. Metric value filtering is also supported here. By combining search filters, you can find relationships between different types of terms.

For example, you can search for your brand in the search term filter, select all brand-related terms, and then input competitor brand names or product categories into the associated search term filter. This helps identify competitor or category terms with high search volume, which can be included in your ad strategy.

The word cloud in the associated search page visualizes specific associated search terms, with word size representing search volume, which can be switched to conversion indicators like purchasing users or sales by changing the metric in the upper-left corner.

The lower section provides a detailed data table of associated searches, with options to hide the word cloud, expand the list to full screen, download the data, and customize columns.

3. Model Customization

The model's custom conditions offer filtering by ad account/ad type/ad campaign, allowing you to analyze data for specific campaigns and supporting Amazon purchase data insights to expand non-advertising reach data. Additionally, you can define customer-purchased ASINs, focusing the model on conversions for selected products.

Compared to Amazon's search term report, AMC data offers ASIN-level reports for browsing, adding to cart, and conversion, showing the search terms consumers used to find a product and all search behavior, not just the keywords clicked in ads.

If keyword testing was insufficient for a potential ASIN due to limited ad placements, making it difficult to identify suitable long-tail keywords, the AMC search term report allows you to see which terms were searched by users who purchased the ASIN. This enables you to quickly identify high-potential long-tail keywords and shorten the testing period significantly.

In custom models, you can also define the lookback period between search and conversion behavior.

Extending the lookback period can reveal more search term data, but longer intervals weaken the causal relationship between search and conversion and may inflate conversion data.

4. Model Data

Like other AMC models, the Search Term Analysis model primarily analyzes user behavior data.

Amazon's backend Search Term reports are based on Amazon Advertising's attribution logic. They focus on ad-driven conversions, such as purchases made within 7 or 14 days after clicking an ad, and are more oriented toward evaluating ad effectiveness.

In contrast, the AMC Search Term Analysis model focuses on the relationship between user search behavior and conversion. It tracks what terms users searched for before making a purchase, regardless of ad clicks or Amazon's attribution logic and period. This helps in-depth understanding and optimization of ad targeting strategies.

For example, if a user searches two terms within one day and completes a purchase, the Amazon search term report attributes the purchase to the ad's last-clicked keyword. In the AMC model, both search terms are recorded as influencing conversion, regardless of whether the user clicked an ad.

From the user's perspective, both search behaviors could potentially influence conversions, and therefore both will be tracked.

Due to AMC's privacy protection, search terms searched by fewer than 100 users are filtered out. This means many long-tail terms with low search volume may be excluded. Pulling multi-month data using a custom model can increase visibility for lesser-searched terms, as search volumes might exceed 100 times. Extending the data period helps access more data for niche search terms.

Due to data filtering, the AMC search term model does not provide advertising cost-related data like CPC or ACOS since filtered cost data becomes distorted and loses reference value.

Glossary