How to Analyze Your Amazon Search Query Performance Report with Claude

Amazon's Search Query Performance report is the most underused data asset most sellers have access to. It tells you exactly which search terms are driving impressions to your products, how many of those impressions result in clicks, and how many clicks convert to purchases - broken down at the query level, for every search term associated with your brand over the reporting period.

The problem is not the data. The problem is the extraction. A properly analysed SQP report contains a prioritised keyword strategy, a conversion optimisation agenda, a PPC campaign expansion plan, and a CTR improvement brief - all sitting in the same CSV file. Getting it out manually requires enough spreadsheet work that most teams do it quarterly at best, or delegate it to an analyst who produces a slide deck that's already three weeks old by the time it's acted on.

This is Part 3 of the Claude x Amazon series. This post is a complete walkthrough: how to export your SQP data, how to prepare it for Claude, the exact prompt structure that produces actionable output, and what the output actually looks like in practice. Part 4 covers where this workflow hits its structural ceiling - and what changes when the data layer is continuous.

What the SQP Report Actually Tells You

The SQP report is a brand-level view. It shows performance for all search queries that resulted in any interaction with your brand's ASINs - impressions, clicks, basket adds, and purchases - compared against the total market for each query.

The key metrics and what they actually mean:

The most valuable analysis compares these metrics against each other, not against benchmarks in isolation. A query with 40% impression share and 8% click share tells a different story than a query with 8% impression share and 8% click share. The first is a CTR problem; the second is a visibility problem. Claude can make that distinction if you give it the right data - a spreadsheet sorted by a single column cannot.

Step 1: Export and Understand Your Data

The SQP report lives in Brand Analytics inside Seller Central. Navigation: Seller Central → Brands → Brand Analytics → Search Query Performance.

Select your brand, set the reporting period to “Monthly,” choose the most recent complete month, and select “Brand view” with “Simple” format. Download as CSV.

A few things to know about the raw export before you do anything with it:

• The first row is a metadata header (brand name, reporting range, date). Row 2 is the column headers. Your data starts at row 3.
• Queries are sorted by Search Query Score - Amazon's composite relevance metric, not by volume. Re-sort by Search Query Volume when looking for priority opportunities.
• Some purchase and basket add cells will be empty (not zero - empty) where counts are too small for Amazon to display without risk of identifying individual buyer behaviour. Treat these as effectively zero for analysis purposes.
• The report covers your full brand, not individual ASINs. If you have products in very different categories under the same brand, the query mix will be broad. Keep this in mind when interpreting results.

Step 2: Prepare Your Data for Claude

The raw CSV has 34 columns. Claude can handle this, but you'll get cleaner output if you reduce it to the columns that matter for the analysis you want to run. For the core SQP analysis, the relevant columns are:

• Search Query
• Search Query Volume
• Impressions: Brand Share %
• Clicks: Brand Count
• Clicks: Brand Share %
• Clicks: Click Rate % (total market CTR for this query - not brand-specific)
• Basket Adds: Brand Share %
• Purchases: Brand Count
• Purchases: Brand Share %
• Purchases: Purchase Rate %

In Excel or Google Sheets: select just these columns, copy, and paste into a new sheet. Delete rows where Search Query Volume is very low (under 50 for most catalogues) - these are statistically thin and add noise to the analysis. Sort by Search Query Volume descending.

For most catalogues, this leaves you with 100 to 400 rows. Claude can handle 400 rows of tabular data comfortably in a single prompt. If your brand has more, focus on the top 300 by volume - these represent the queries where your decisions will have the most impact.

Copy the reduced table as a tab-separated or comma-separated block, ready to paste into the prompt below.

Step 3: The Analysis Prompt

This is the core SQP analysis prompt from Part 2 of this series, with additional detail on the output format:

Run this prompt and wait. For 300 rows of data, Claude typically takes 20 to 40 seconds to produce output. The output will be structured in the four sections you defined - no reformatting required.

Step 4: Turning Claude's Output Into Actions

The output gives you three distinct action queues. Here's how to route them:

CTR gap fixes go to listing content (title, main image, primary bullet) and pricing review. If the gap is image-related - Claude will usually identify this when your impression share is strong but click share is weak - this becomes a main image test. If it's title-related, use the listing audit prompt from Part 2 to generate the rewrite.

Conversion gap fixes go to deeper listing content (bullets 3 to 5, description, A+ Content), review response strategy, and pricing analysis. Conversion gaps often reveal missing information - buyers are interested enough to click, but the listing doesn't answer the question that would close the purchase. The review intelligence prompt from Part 2, run against reviews for the affected ASINs, often surfaces exactly what that missing answer is.

PPC harvest candidates go directly to your campaigns team: these are search terms performing well organically that need exact match protection before competitors start bidding aggressively on them. This is one of the clearest immediate ROI actions from an SQP analysis - and one that most teams miss because manual SQP review rarely surfaces it efficiently.

What the Output Actually Looks Like

Here's a condensed example of the kind of output this prompt produces (synthesised from real analyses, with brand-specific data removed):

Section 1 - CTR Gap (one example):

This is not a keyword tool output. It's a specific, reasoned diagnosis with a specific action - the kind of analysis that previously required an analyst who understood both the data and the listing to produce together.

The Limits of This Workflow

This workflow is genuinely useful. It also has limits worth naming before you try to scale it.

The data is always a month old. The SQP report covers a completed month. By the time you export, prepare, and analyse, you're working with data that's four to six weeks behind reality. For a category with seasonal demand or fast-moving competitors, that lag is meaningful.

Claude has no memory. Next month's analysis starts blank. There's no way to ask whether the CTR fix implemented last month actually improved click share for the targeted query - because Claude has no record of what was implemented. Trend detection requires you to run the analysis twice, compare outputs manually, and infer the delta yourself.

Implementation is not connected. Claude produces recommendations. Acting on them - rewriting the title, updating the main image brief, harvesting the PPC terms - requires a separate manual workflow with no tracking of what was implemented or what effect it had.

It does not scale across portfolios. Running this analysis for one brand in one market takes thirty to forty-five minutes including export, preparation, and output review. For five brands across three markets, that's a part-time job - before any implementation work begins.

These are the structural failure modes of the manual Claude workflow covered in Part 4 of this series. If you're running one brand or testing the approach for the first time, the workflow above is a solid starting point. If you're managing a serious catalogue, the ceiling appears quickly.

Frequently Asked Questions

Where exactly do I find the SQP report in Seller Central?

Seller Central → Brands → Brand Analytics → Search Query Performance. You need to be Brand Registered to access it. If you don't see the “Brands” menu item, your account either isn't Brand Registered or the brand hasn't been linked correctly. Use the brand-level view with monthly granularity for the workflow described in this article.

How often is the Amazon SQP report updated?

The SQP report is available at monthly granularity only - weekly and daily views are not currently available for the brand-level report. New monthly data typically becomes accessible around the 10th-15th of the following month. This means your most recent complete month is always 10-45 days old by the time you can access it.

What exactly does “Brand Share” mean in the SQP report?

Brand Share is calculated relative to all sellers whose ASINs appeared for that search query. Your brand's impression share is the percentage of all search result impressions where one of your ASINs appeared. Your brand's click share is the percentage of all clicks on search results for that query that went to one of your ASINs. High impression share but low click share signals a CTR problem rather than a visibility problem.

Can I run this SQP analysis for multiple Amazon marketplaces at once?

Each marketplace has its own SQP report. Amazon.de, Amazon.co.uk, Amazon.fr, and Amazon.it each generate separate reports with market-specific search behaviour. Run the analysis per-marketplace and treat each as a separate optimisation context - the variation in query patterns between markets is often large enough that insights from one market don't directly apply to another.

What should I act on first after getting Claude's SQP analysis output?

Focus on CTR gaps first - queries where you're already getting impressions but losing clicks. The fix (typically main image test or title refinement) is faster to implement than conversion improvements, and the impact is immediate since you're already visible. Second priority: PPC harvest - high-converting organic terms to move to exact match campaigns. Conversion gaps typically require deeper listing changes and take longer to show results.