A seller prompt workbook built around report denominators, a spend identity and a market-share question the supplied files cannot answer.
The answer that should not be a percentage
Ask what share of the entire U.S. coffee market your brand won last week. Owned seller sales provide a possible numerator only after defining products, channel, geography and sales basis. One ASIN's Search Query Performance provides query-specific funnel counts, not total category sales. Neither file supplies the independent market-wide denominator. The honest answer is 'not measurable from these files,' followed by the missing category definition and source. A neat-looking percentage without that denominator can travel into a budget deck before anyone checks it.
Make the first prompt about the connection, not the answer: ‘Using 1stPage, list the accounts available to my user. For the seller account I select, show the US marketplace and available analytics tools.’ Then give Claude one report question. Ask for seller_analytics when the question is about ASIN sales; use advertising_analytics for paid clicks only if the Ads capability is available. A denied tool call is not a data point.
This is a useful first test for Claude: refusing unsupported precision matters more than fluent narration. Attach only approved data and keep the filename or tool-result ID in the output. Amazon's Seller Assistant plugin announcement describes a beta for sellers in U.S. stores; it does not establish a given account's eligibility or the plugin's exact report coverage. These prompts work with a verified export route, irrespective of whether that beta is available.
Can the attached files support a category market-share percentage for last week? State the needed numerator and independent market-total denominator, including geography and sales basis. If that denominator is absent, answer 'not measurable from these files' and give no percentage.
Put the sales loss on specific ASINs
For seller sales, use an authorized Sales and Traffic report or comparable Seller Sales and Traffic Data Kiosk extract. Keep marketplace, currency, ASIN population and reporting time zone fixed across two completed seven-day periods. Name the metric: ordered product sales are not settlement proceeds or ad-attributed sales. Check the source total before explaining traffic, especially when new or missing ASIN rows change the apparent population. Synthetic hand-check: sales go from $1,000 to $850. ASIN X moves from $400 to $270, while all other ASINs move from $600 to $580. The changes are -$130 and -$20, totaling -$150. X explains $130 of the net $150 decline, not the whole decline. Reconcile ASIN differences with the report total. Sessions at ASIN grain must not be added into an account-wide session total without checking the report's definition. These amounts are arithmetic examples, not observed seller performance.
Using the attached seller Sales and Traffic extract, compare two named complete weeks for one marketplace and currency. Give each ASIN's sales delta and reconcile the sum to the source total. For the largest contributors, name the available sessions and order or unit conversion field, including its numerator and denominator. Mark unavailable fields unknown. Do not attribute the loss to ads.
Make ad spend pass an identity check
Amazon Ads reports require their own approved access. A seller Sales and Traffic export cannot tell you that CPC rose. With a campaign export containing clicks and spend, calculate CPC as spend divided by positive clicks, preserving campaign IDs, currency, completed dates and report settings. An ad-attributed sales field has its own attribution window; it is not interchangeable with total ordered product sales.
Synthetic check: 1,000 clicks at $0.80 cost $800. Another completed period has 1,200 clicks at $0.90, costing $1,080. The $280 rise splits into $160 for 200 additional clicks at the old CPC and $120 for the $0.10 CPC increase across 1,200 new-period clicks.
That is an accounting identity, not proof that bids changed. Verify reported spend against clicks times CPC with rounding tolerance. Zero-click campaigns need separate treatment rather than division by zero.
For matching campaigns and completed dates, show clicks, spend and calculated CPC in each period. Check spend against clicks times CPC, then decompose the spend difference as (new clicks - old clicks) × old CPC plus new clicks × (new CPC - old CPC). Flag missing campaigns, zero clicks, currency differences and incompatible attribution settings.
Eight days of cover is not a delivery date
An inventory snapshot records a time, not a promise about tomorrow. Join seller SKU and ASIN at a compatible grain in one marketplace, and keep FBA stock separate from merchant-fulfilled quantity. Inbound units are not currently sellable. A simple conditional cover measure divides sellable units at the snapshot by average daily units ordered over a stated window. Forty sellable units divided by five daily units gives eight days at the observed rate, assuming no new supply and unchanged demand. It does not forecast the calendar day of a stockout.
Check the numerator in the inventory report and the denominator in the seller sales extract. A zero-demand denominator produces no finite cover figure; missing inbound timing leaves replenishment unknown. A source that only contains today's stock cannot prove the listing was unavailable last Tuesday. The prompt should keep those unknowns visible rather than supplying an automatic reorder quantity.
For matched seller SKU and ASIN records, report sellable units at the inventory snapshot, units ordered in the prior 28 complete days, average daily units and conditional days of cover. Split FBA and merchant-fulfilled stock. Mark zero-demand rows, missing inventory and unknown inbound dates without assigning a stockout date or reorder amount.
Eighty clicks out of a thousand, but of what?
Amazon documents Search Query Performance as a requested JSON report for eligible Brand Registry sellers with the Brand Analytics SP-API role. Its valid period boundaries are WEEK, MONTH and QUARTER. The report distinguishes total-query impressions, clicks, cart adds and purchases from an ASIN's counts and shares. Sales and Traffic permission alone does not confer access. Match query, ASIN, store and completed period before comparing shares.
For a synthetic query with 1,000 total query clicks and 80 clicks for the ASIN, the click share is 8% for that query and period. It is not category market share, organic rank or a competitor-sales estimate. Recompute one reported share from its counts, then look for missing funnel stages and absent rows. An absent ASIN/query row is not evidence of zero demand. Keeping the denominator beside every percentage stops an attractive but unsupported claim at the point of calculation.
For one ASIN and a completed Search Query Performance period, show the query and available total and ASIN counts at each funnel stage. Name each share's denominator and recalculate click share from counts. Mark missing stages or rows; do not infer organic rank or category sales.
Sources and boundaries
Check the current permissions and report definitions before acting on an answer. Example prompts are questions, not live account results.