Amazon BI / Commercial Diagnosis
Market Share Loss Diagnosis Framework
How to explain a product decline without confusing correlation with cause: separate visibility, click share, conversion, price, inventory, reviews, competition, and advertising before choosing an intervention.
Revenue is an outcome. It is not a diagnosis.
A sales decline can be produced by fewer impressions, weaker click-through rate, lower conversion, a smaller average selling price, lost availability, or a combination of them. Market-share loss adds another layer: the product may be declining because the category contracted, or it may be losing position inside a stable or expanding category.
Revenue = category demand × impression share × click rate × conversion rate × average selling priceThis decomposition forces each explanation to name a mechanism and an observable signal. It also prevents the common mistake of changing copy, bids, or price simply because those levers are available.
Locate the break before explaining it.
Are impressions or query share down?
Check indexation, organic rank, sponsored coverage, category position, stock, and suppression.
Are shoppers clicking less?
Compare main image, title, price, coupon, rating, delivery promise, badges, and new SERP competitors.
Are detail-page visits converting less?
Review offer stability, reviews, returns, variation clarity, product communication, and competitive substitutes.
Is revenue stable but profit weaker?
Reconcile price, fees, landed cost, ad spend, promotions, returns, and product mix.
Was demand interrupted?
Trace stockouts, inbound delays, Buy Box loss, fulfillment mode, delivery speed, and stranded inventory.
Did the category move?
Use search-volume trends, competitor BSR, pricing, promotions, review velocity, and share metrics.
A 50% decline with stable rank and improving returns
One archived monthly audit showed units falling from 4,105 to 2,049 and revenue from roughly $73.6K to $36.8K. Contribution margin fell more sharply, from approximately $14.5K to $5.3K. A superficial reading could blame content quality or returns. The diagnostic evidence did not support that conclusion.
| Tested explanation | Evidence | Conclusion |
|---|---|---|
| Out of stock | The offer was available; no material stock interruption was documented. | Not supported for the reviewed period. |
| Fee shock | The FBA fee movement was approximately $0.02 per unit. | Too small to explain the margin loss. |
| Product dissatisfaction | Return rate improved from 1.05% to 0.63%. | Returns moved in the opposite direction. |
| Keyword visibility | Four reviewed priority terms retained top organic positions; several also held top sponsored positions. | No broad ranking collapse in the sampled terms. |
| Advertising efficiency | ACoS remained near the operating threshold, but individual exact targets showed room for bid reduction or removal. | Selective optimization, not indiscriminate cuts. |
Evidence boundary: figures are rounded and the product/client identity is omitted. This monthly audit did not contain complete category-share history, so the defensible conclusion is that several common causes were ruled out—not that one remaining factor was proven.
Share metrics tell a different story at each funnel stage.
A separate anonymized August 2024 Search Query Performance export covered 1,000 queries. Weighted share is calculated as the summed brand count divided by the summed marketplace count at each stage—not as an unweighted average of row percentages. The brand held 2.54% impression share, 3.19% click share, 3.24% cart-add share, and 3.72% purchase share.
Weighted brand share by marketplace funnel stage
Percent of marketplace stage count; August 2024 brand-view export; n=1,000 queries. Bars share a zero baseline and a 4% maximum.
Source counts: 10,257,972 of 403,150,568 impressions; 219,562 of 6,880,407 clicks; 58,977 of 1,821,500 cart adds; and 27,458 of 737,810 purchases.
Because share increased deeper in the funnel, the aggregate pattern is more consistent with a visibility constraint than with weak downstream conversion. It is still descriptive: query mix, price, product mix, and traffic sources can differ by stage, so the chart does not estimate incremental advertising impact.
Visibility gap
When purchase share exceeds impression share, investigate where relevant query coverage or rank is insufficient before rewriting the detail page.
Conversion gap
When impression and click share are healthy but purchase share falls behind, prioritize offer, content, reviews, price, and fulfillment.
This brand-level report is separate from the product audit above. It demonstrates the comparison method and must not be merged with the product snapshot as if both describe the same ASIN or period.
End with a ranked explanation, not a dashboard tour.
| Output | Required content |
|---|---|
| Observed change | Magnitude, period, baseline, and whether category demand also moved. |
| Supported drivers | Evidence, mechanism, estimated contribution, and confidence level. |
| Ruled-out drivers | What was checked and why it cannot explain the material change. |
| Unknowns | Missing reports, unresolved timing, and tests needed to reduce uncertainty. |
| Action plan | Owner, lever, expected mechanism, guardrail, and review date. |
The Amazon operating handbook reinforces this logic: mature-product analysis should track revenue, contribution margin, impressions, and impression share over time, then separate impression loss from conversion loss before acting.
Move from symptom to mechanism in a fixed order.
1. Define the comparison before reading the delta.
Start by naming the product scope, marketplace, calendar periods, and currency. A month-over-month decline can reflect seasonality, a different number of selling days, a promotion ending, or a category-wide contraction. The percentage formula is simple—current period minus prior period, divided by the prior period—but its meaning depends on whether the periods are genuinely comparable.
2. Separate volume from unit economics.
In the archived product snapshot, units and revenue each fell 50.1%, while contribution margin fell 63.6%. Similar unit and revenue declines suggest average revenue per unit did not drive the headline sales loss. The larger margin decline says economics weakened in addition to volume. Calculating margin rate makes that visible: approximately $14,529.98 divided by $73,631.43 equals 19.7%, while $5,281.59 divided by $36,774.97 equals 14.4%. The 5.4-point decline is a percentage-point change, not a 5.4% relative decline.
3. Test hard operational breaks first.
Availability, suppression, Buy Box ownership, and delivery promise can interrupt the funnel before shoppers evaluate content. Because the reviewed offer remained available and the documented fee movement was only about $0.02 per unit, neither observation could plausibly explain the full magnitude by itself. “Not supported” does not mean “impossible”; it means the available record did not show a material interruption during the reviewed window.
4. Read funnel shares from top to bottom.
The separate brand-view export asks a different question. Weighted impression share of 2.54% means the brand generated 10,257,972 of the 403,150,568 marketplace impressions represented by those 1,000 queries. Purchase share of 3.72% means 27,458 of 737,810 purchases. Because the denominator changes at every stage, these shares describe relative representation within each stage; they are not conversion rates. The rise from impression to purchase share suggests the brand performed better among shoppers who encountered it than its initial visibility would imply.
Decision implication: protect working conversion assets while investigating query eligibility, rank, paid coverage, category demand, and competitor exposure. A wholesale content rewrite is a weak first response when downstream share is stronger than impression share.
5. Rank explanations by confidence.
Use three labels: supported, ruled down, and unresolved. Stable sampled keyword positions rule down a broad ranking-collapse story, but four priority terms cannot represent the entire query portfolio. Improving returns rule down dissatisfaction as the main explanation, but return rate is a lagging and incomplete signal. Missing category history leaves demand shift, query-mix change, competitor expansion, and traffic-source change unresolved.
What to collect, compare, and decide next.
Scope
Freeze ASINs, marketplace, period, currency, fulfillment mode, and comparison baseline. Annotate promotions, stock events, and catalog changes.
Demand
Compare category search volume, query impressions, impression share, competitor rank, seasonality, and external traffic before calling the issue “share loss.”
Funnel
Reconcile impressions, clicks, click-through rate, sessions, conversion, units, and average selling price. Identify the earliest stage where performance diverges.
Economics
Bridge revenue to contribution margin through price, discounts, fees, landed cost, ad spend, returns, reimbursements, and product mix.
Alternatives
Document competing explanations and the evidence that would confirm or weaken each one. Do not turn missing data into a confident narrative.
Action
Choose the smallest reversible intervention tied to the suspected mechanism, assign an owner and review date, and protect a guardrail metric.
A useful final recommendation has the form: “We observed X, evidence Y supports mechanism Z with this confidence, alternatives A and B remain open, and action C should move metric D without harming guardrail E.” That structure makes the analysis auditable and prevents a dashboard from masquerading as a decision.