Skip to main content
All case studies

Investigating a sales drop: search demand and advertising

When sales fall, should the team change the offer, spend more on advertising, or look for a change in how shoppers reach the products? This historical analysis by the DataBrill team began with that question for a US Amazon store. Changing the listing before understanding the traffic would have meant acting on an explanation the team had not yet checked.

The figures come from the original analysis. The seller is described as a kitchen and household seller; products and search terms have been changed to protect its identity. The trash-can labels below are substitutions, not a separate measurement of that market.

Put the reports on the same timeline

The team compared store orders, Sponsored Products advertising reports, and Amazon Brand Analytics search frequency ranks. Each answered a different part of the question: what the store sold, what happened to paid traffic, and how frequently relevant searches ranked relative to other searches.

The comparison used daily medians for the 28 days before May 3, 2023, and the 14 days beginning on that date. Keeping the windows explicit matters because sales were already drifting down before May 3. A difference between these windows includes that earlier movement; it cannot all be assigned to an event on a single day.

Lower sales and ad clicks; little change in ad response
Daily store sales · USD
Before$9,743
After$8,424
Ad clicks per day
Before1,944
After1,529
Ad click-through rate
Before0.56%
After0.55%
Ad conversion · orders per click
Before10.8%
After11.0%
Search frequency rank · “trash can with lid”Before 6,366After 56,407Ordinal position: lower is more frequent. This is not search volume.
Daily medians, US store: before = April 5–May 2, 2023 (28 days); after = May 3–16 (14 days). Each bar pair has its own scale, starting at zero; lengths cannot be compared between measures. Advertising covers Sponsored Products with seven-day order attribution, not all store traffic. Sales were already declining; this is not an overnight or isolated causal effect.

What changed around May 3

Measure Before After
Daily store sales, USD $9,743 $8,424
Search frequency rank, “trash can with lid” 6,366 56,407
Ad clicks per day 1,944 1,529
Ad click-through rate 0.56% 0.55%
Ad conversion, orders per click 10.8% 11.0%

Daily medians, US store, April 5-May 2 versus May 3-16, 2023. Sales cover the whole store. Advertising measures cover Sponsored Products, with seven-day attribution for orders; they do not measure all store traffic.

Search frequency rank is an ordering, with a lower number indicating a more frequent search. The move from 6,366 to 56,407 shows a large change in relative position, but dividing those numbers does not tell us how many shoppers were lost. Brand Analytics rank is not a search-volume count.

Store sales were roughly 14% lower across the two median windows. That is a description of the comparison, not a 14% overnight drop or an estimate of the effect of search suggestions alone.

Why the team looked at traffic

Ad clicks fell while median ad click-through and conversion rates changed little. Click-through rate is clicks divided by impressions; the conversion measure here is attributed ad orders divided by ad clicks. The distinction directed attention toward exposure: fewer clicks could account for lower sales without requiring a large change in how ad shoppers responded once they saw or clicked the offer.

Ellis confirmed that the team had not changed bids or budgets in early May. That reduced the case for an intentional advertising cut as the explanation. It did not establish that every part of the product page, price, or offer was faultless.

The team’s explanation was that Amazon had changed its search suggestions, affecting relevant search terms. The team identified this in August, about three months after the May change. No autocomplete history had been collected, so the suggestion change is a documented team observation and interpretation rather than a measured time series.

A rank change alone could also reflect a change in how searches were counted. The store’s own ad clicks and sales changed independently of Brand Analytics, which supported investigating a real traffic change. Those records still could not isolate how much of the decline came from suggestions, and other causes may have contributed.

What the evidence could not settle

The available 2023 history did not include store-wide sessions or organic conversion. Stable-looking ad conversion therefore says something about ad shoppers, not about the conversion of every visitor to the store. Treating it as a whole-store measure would make the conclusion stronger than the evidence.

The search rank partially rebounded later, in June, without a matching recovery in the lost-target clicks and sales examined in the analysis. That also limits a simple explanation in which one search measure determines everything else. This was an investigation across reports, with no controlled experiment or isolated causal estimate.

Use the diagnosis to choose the work

The team reviewed alternative relevant search terms and adjusted its attention accordingly. Understanding the change also allowed it to concentrate on other products and launches. The records do not supply a recovery figure to attach to that decision.

The useful result was a more specific question about traffic and the work that followed from it. For a way to make that comparison in your own reporting, read Why did your Amazon sales change?. The sustained-growth analysis follows the opposite question: whether a higher sales level held. Core’s platform explains the data pipeline behind keeping Amazon reports available for analysis.