Why did your Amazon sales change?
A fall in sales does not tell you whether to rewrite a listing, change a price, or adjust advertising. A rise does not tell you which decision to repeat. Before choosing an action, the useful question is which part of the buying process changed and which records could distinguish the possible explanations.
Start with a comparison that a teammate can review: the same marketplace and product scope, clearly named periods, and the actions the team took during them. Then work from the sales total toward the traffic and offer measures that could explain it. A missing report is a limit on the answer, rather than a reason to assume its measures held steady.
Investigate placement and the offer alongside demand. These paths suggest what to compare; they do not isolate a cause.
Start with units and price
Revenue equals units sold multiplied by the realized average price per unit, when all three use the same scope and revenue basis. That calculation separates two movements that a sales chart combines. A store can sell more units and report less revenue if its average realized price falls far enough.
The average also reflects product mix and discounts. If more sales come from a cheaper pack, the store’s average price can fall even when no listed price changed. Compare like products as well as the total before deciding that a price change explains the result.
This is arithmetic decomposition. It shows where to look next, but it cannot establish why units or prices changed.
Trace how shoppers reached the offer
Demand and visibility answer different questions. Demand concerns shoppers searching for relevant products; visibility concerns whether your offer appears where those shoppers can encounter it. A store can lose traffic because the searches changed, because its placement changed, or through a combination of influences.
Advertising and organic search both have a path through impressions, clicks, and purchases. Click-through rate (CTR) is clicks divided by impressions. Conversion rate (CR) needs an explicit definition: ad orders per click, units per click, and units per session are different measures. A multi-unit order is one reason not to treat them as interchangeable.
Costs belong in the comparison too. Cost per click (CPC) describes advertising spend per click. More sales may arrive with a different advertising bill or a different price, so a sales increase alone does not establish a profit gain.
Choose the reports for the question
| Question | Records to compare | Limit |
|---|---|---|
| Did units or realized price change? | Orders, quantities, and sales amounts | Keep products, marketplace, and revenue basis consistent |
| Did relevant search demand change? | Brand Analytics search frequency rank and Search Query Performance counts | Rank is ordinal; query counts cover the searches examined |
| Did paid exposure or response change? | Sponsored Products impressions, clicks, and attributed orders | Advertising scope and attribution differ from store-wide traffic |
| Did store traffic or units per session change? | Sales and Traffic sessions and units | Units per session is not the share of unique visitors who ordered |
These report roles describe how to choose evidence, not a promise that every historical series will be available. Check coverage before making the comparison.
A search frequency rank tells you relative order among searches. Its ratio cannot be converted into a ratio of search volume. Search Query Performance supplies counts for the queries being examined, but a sample of queries does not become the entire market merely because its data is detailed.
The traffic reports also need their own definitions. Attributed ad orders follow an advertising attribution window; store sessions measure something else. Align the periods and product coverage, and keep the difference visible when comparing them.
Separate a calculation from an explanation
CTR is calculated from clicks and impressions. Advertising spend is clicks multiplied by CPC when the measures cover the same scope and period. Those relationships are useful checks on a comparison, but strong agreement between calculated measures is not independent evidence of a business cause.
Other relationships are plausible influences to investigate. A different price can change how shoppers respond to an offer, while advertising bids can affect costs and placement. Their effect in a particular case needs evidence; neither supplies a universal instruction to lower prices or raise bids.
Put the team’s actions on the timeline beside the observed changes. Check whether the movement began earlier, how the same season looked in the comparison period, and whether several things changed together. Correlation can reflect shared timing or arithmetic rather than the mechanism you are trying to establish.
Before accepting an explanation, check report freshness, attribution, marketplace, and product scope. A comparison that changes those definitions midway may produce a precise-looking answer to a different question.
Read a decline and a rise differently
In the sales-decline analysis, falling ad clicks alongside little change in median ad response rates concentrated the investigation on traffic. The missing store-wide conversion history still limited what could be said about the offer as a whole. The team’s search-suggestion explanation was not a measured autocomplete history.
The sustained-growth analysis followed a deliberate price change, later search positions, and unit sales over a longer period. The sequence was informative, while competitor changes, advertising, and seasonality remained part of the explanation. A higher level that held could be observed without claiming to have proved Amazon’s ranking mechanism.
Both cases used several reports to narrow the question. They did not turn a sales chart into a complete causal answer.
Write the next investigation before making the change
A useful analysis ends with the evidence, the interpretation, what is still uncertain, and the next record or comparison that could change the decision. If the evidence points toward lost exposure, name the placements or searches to inspect before commissioning a listing rewrite. If units rose after a price change, check the economics and competing influences before treating the price as a rule to repeat.
The records should follow the business question. Inbound reimbursement reconciliation needs shipment, receipt, and payment records rather than a sales-traffic comparison. The same discipline applies: define the scope, preserve the matching rules, and make the remaining uncertainty visible.
From Amazon reports to a database your team can use explains how to keep those records available for repeat analysis. Core’s platform describes its role in collecting Amazon data into a database your team can query.