Availability, Stock-Outs and the Signal They Create
The most common cause of a rank pattern that looks like a market event. Capturing availability alongside rank resolves it, and almost nobody does.
A product that cannot be bought does not sell, and its rank decays exactly as if demand had collapsed. This produces the single most convincing false conclusion available from rank data.
The signature
A smooth decay as the recency weighting on past sales fades with no new sales replacing them.
Then a sharp recovery to approximately the previous level within days of restocking.
The recovery is the tell. Genuine demand decline does not rebound cleanly to its prior level in a few days.
Duration matches the outage, which is usually days to weeks rather than the months a real decline takes.
Why it matters so much
For competitive analysis, a competitor stock-out reads as a product failing. Acting on that — entering a category, adjusting price, reallocating spend — on the basis of a temporary supply problem is a real and common error.
For your own analysis, an outage during a campaign period destroys the measurement, and the campaign is judged on a period when the product could not be bought.
For market sizing, products unavailable during the observation window are undercounted.
Capturing it
Record availability with every rank observation. In stock, out of stock, limited, backordered, or unavailable.
Record the seller too where the marketplace shows one. A product available from a third party but not the primary seller behaves differently.
Record delivery timing where shown, since a long lead time suppresses conversion without being a stock-out.
These fields are cheap and they cannot be added retrospectively, which is the recurring theme of context capture.
Analysing around outages
Exclude outage periods from baseline and comparison calculations rather than averaging through them.
State the exclusion. "Twenty-two of thirty days, excluding an eight-day availability gap."
Do not interpolate across an outage. The values were not merely unobserved; they were depressed by a known cause.
When comparing two products, restrict to the period both were available, or the comparison is between a product and a supply problem.
What an outage tells you that is real
Demand exceeded supply, which is information about the product and about the seller's operations.
Repeated outages indicate a supply chain that cannot support the demand, which is a competitive weakness worth knowing about.
The recovery slope shows how quickly the audience returns. A product that recovers immediately had waiting demand; one that recovers slowly lost customers to substitutes.
Timing matters. An outage during a peak season is far more damaging than the same outage in a quiet month, and the rank data shows the difference.
Partial availability
Harder to detect and worth knowing about.
Variant-level outages. A parent listing showing available while the popular size is not.
Regional availability differences on marketplaces that serve multiple regions.
Buy box loss, where the product is available but not from the seller you are tracking, which suppresses their sales while the listing appears healthy.
These produce partial rank decay that is harder to distinguish from genuine decline, and they are the reason seller-level capture matters for serious competitive work.
The check before any conclusion
For any unexplained rank decline in a tracked product:
Was it available throughout? If unknown, the analysis is provisional.
Did it recover sharply? If so, supply rather than demand.
Did the basket move? If so, market rather than product.
Did price change? If so, price rather than demand.
Four questions, and they account for most rank movements that are otherwise attributed to something interesting. Running them first is the difference between analysis and storytelling.
Logging outages for later analysis
An outage log turns a nuisance into a research asset.
Record for each outage: product, start, end, and whether it was full or partial.
Detected automatically from the availability field, with a minimum duration to avoid noise.
Reviewed weekly to confirm the automatic detection matches reality.
The log enables the substitution study described elsewhere: which competitors gained during each outage, and by how much relative to the basket.
It also enables clean analysis, because outage periods can be excluded systematically rather than being noticed ad hoc.
And it is a competitive finding in itself. A competitor with frequent outages has a supply constraint, which is durable information about their operation that is hard to obtain any other way.