Executive answer: the data hierarchy that matters most
If you want a replenishment forecast that is actually useful, start with the most recent clean sales window, then layer current stock position, reserved units, inbound supply and lead times, and only then adjust for promotions, pricing, PPC, seasonality and launches. In practice, that means I would trust recent uninterrupted sales velocity first, not a blended average that includes stock-out distortion or stale history. Amazon data is useful, but it is not a single source of truth; you have to reconcile reports before you act on them. Amazon’s inventory reports are snapshots and may be out of date almost as soon as they are requested, so the forecast should be a planning view, not a blind acceptance of platform numbers.
Start with demand: recent sales velocity and clean historical windows
For a living ASIN, the cleanest demand signal is the most recent uninterrupted sales window. If the product has not been interrupted by stock-outs, the recent 30 days is often the best practical baseline because it is close to current market reality and less likely to be polluted by older pricing, promo or traffic conditions. If the ASIN has been out of stock, I would not average across the outage and the recovery period; I would go back to the last clean 30-day window before the stock-out and use that as the base, then adjust for known changes. That is a practitioner rule, not an Amazon policy, but it is usually a better planning input than a raw historic average. Amazon’s inventory reports are only a spot check of saleable inventory, so sales history and inventory state should be read together rather than in isolation. ([sellercentral.amazon.com](https://sellercentral.amazon.com/help/hub/reference/external/G561?locale=en-US&utm_source=openai))
The reason averages mislead is simple: stock-outs suppress recorded sales, and post-stock-out recovery often reflects a different ranking and traffic state from the one you had before. If you are planning for a hero ASIN, the forecast should reflect actual sell-through in the last clean period, not a distorted average that includes unavailable inventory. In founder-led planning, I would also separate by SKU or brand where the account structure demands it, because multi-brand or multi-market activity can hide very different demand patterns. That separation is a judgement call, but it is usually the difference between a forecast that looks tidy and one that is operationally useful.
Add supply: on-hand, reserved, inbound, stranded and lead time data
A replenishment forecast is not just a demand model; it is a supply position model. The core inputs are on-hand inventory, reserved stock, inbound shipments, unfulfillable or stranded units, and the lead time from purchase order to sellable stock. Amazon’s FBA inventory view separates available, inbound, unfulfillable and reserved stock, and the reserved inventory report breaks reserved units into reasons such as customer orders, staging and FC processing. Those categories matter because they change what is actually available to sell now and what is merely still in the pipeline. ([sellercentral.amazon.com](https://sellercentral.amazon.com/help/hub/reference/external/201972720?locale=en-US&utm_source=openai))
In practice, the most common planning mistake is to treat inbound as if it is already available. It is not. If goods are in transit, processing or reserved, they may not support the next demand spike on the date you need them. That is why lead times belong in the forecast itself rather than in a separate operations note. I would also keep stranded or unfulfillable stock visible, because inventory that cannot be sold still affects working capital and can hide the real gap between demand and supply. Amazon’s inventory pages are useful for this, but they still need reconciling against your own stock records before you buy more. ([sellercentral.amazon.com](https://sellercentral.amazon.com/help/hub/reference/external/G561?locale=en-US&utm_source=openai))
Then adjust for demand-shaping events: promos, pricing, PPC, seasonality, launches
Once you have a clean demand baseline and a true supply picture, adjust for the things that can move demand faster than history alone can explain. Promotions, price changes, paid media, seasonality and launches all deserve explicit treatment. Amazon Ads guidance shows that search term reports, targeting reports and related campaign reports can help identify the search terms and products that are generating clicks and sales, which makes them useful for spotting demand-shaping activity rather than just measuring ad performance. ([advertising.amazon.com](https://advertising.amazon.com/help/G3HEFZYWZF84NPS9?utm_source=openai))
For seasonal or gifting brands, I would not smooth away the peaks. If Valentine’s Day, Father’s Day or Q4 is a real part of your trading pattern, the forecast should say so. The same applies to launches and traffic pushes: if paid media is about to lift demand, the forecast must reflect that uplift before the stock lands, not after. Amazon’s own guidance on restock recommendations and related tools is useful context, but the brand still needs to know whether a planned campaign, price move or marketplace push will make the next period materially different from the last one.
Which Amazon reports to trust first, and what each one is good for
| Report or page | Best use in replenishment planning | Watch-outs |
|---|---|---|
| Inventory reports | Quick snapshot of current listing and inventory position | Snapshot only; can be stale almost immediately and must be reconciled |
| Amazon-fulfilled inventory / FBA inventory pages | See available, inbound, reserved and unfulfillable stock states | Inbound is not yet sellable; reserved units need interpretation |
| Reserved Inventory report | Break down units tied up in customer orders, staging or FC processing | Reserved does not mean available for immediate sale |
| Business reports and sales by ASIN / child variation exports | Build the sales velocity baseline and clean demand window | Use with care if stock-outs, returns or channel overlap distort the trend |
| Amazon Ads search term and targeting reports | Understand whether advertising activity is shaping demand | Ad reports are not a substitute for sell-through and stock planning |
The practical point is that no single report is enough. Amazon’s own documentation supports using multiple report types, but the job of the operator is to reconcile them into one planning view. That is especially important when sales, inventory and advertising are being read by different teams. The more fragmented the account, the more important it is to export, join and sanity-check the data before you commit to a purchase order. ([sellercentral.amazon.com](https://sellercentral.amazon.com/help/hub/reference/external/G561?locale=en-US&utm_source=openai))
Common forecast errors: stale data, stock-outs, returns, duplicated counts
The most common errors are also the most expensive. First, stale data makes a forecast look correct on paper even though the underlying position has already moved. Second, stock-outs make historical averages lie. Third, returns, cancellations and one-off anomalies can inflate or suppress the apparent trend if you do not isolate them from underlying demand. Fourth, double counting across channels or systems can make inventory or demand look healthier than it really is. Amazon explicitly warns that inventory reports are snapshots and should be reconciled with your own records, which is exactly the right mindset here. ([sellercentral.amazon.com](https://sellercentral.amazon.com/help/hub/reference/external/G561?locale=en-US&utm_source=openai))
From an operator’s point of view, this is where the hygiene work matters. If a PPC team is scaling spend from one version of the truth while the ops team is looking at another, you can end up paying to rank a listing that is about to stock out. Equally, if returns or cancellations are left in the trend without adjustment, you may think demand is soft when the problem is actually data quality. The best forecast is not the most complicated one; it is the one with the fewest hidden distortions.
When to override Amazon’s forecast with brand-side judgment
I would override Amazon’s forecast whenever the brand-side view is materially more current or more informed. That includes situations where internal sales trends clearly differ from the platform view, where a commercial event is known in advance, or where supply risk is high because the supplier is volatile or the lead time is uncertain. In those cases, the Amazon forecast is still a useful reference point, but it should not be the final decision. This is practitioner judgement, not Amazon policy. The point is to protect service, cash and margin at the same time rather than blindly optimising one number. ([sellercentral.amazon.com](https://sellercentral.amazon.com/help/hub/reference/external/G561?locale=en-US&utm_source=openai))
When working with brands, I usually think in terms of risk allocation. Some ASINs need rank protection, some need cash protection, and some need margin discipline first. A human review should override the platform view when the forecast cannot see the commercial context: a launch, a promotion, a media spike, a supplier delay or a seasonally narrow selling window. For newer or faster-moving brands, that review step is not optional; it is the control that stops a forecast from becoming an automated mistake.
International brands: what to add for multi-market forecasting
If you sell across marketplaces, do not assume one country’s pattern transfers cleanly into another. Treat each marketplace as its own demand window and its own supply chain context. Keep marketplace-specific data separate where behaviour, fulfilment timing or ad performance differ, and only combine it after you have checked that the operating assumptions actually match. That is especially important for promotions, lead times and media calendars, which rarely line up perfectly across countries. Amazon’s reporting and ad tools are available by marketplace and region, but their names and exact labels can vary, so confirm the current interface in the relevant market before you publish a planning workflow. ([sellercentral.amazon.com](https://sellercentral.amazon.com/help/hub/reference/external/G561?locale=en-US&utm_source=openai))
For international brands, I would also add local trading events, import timing and country-specific supply constraints to the core model. A global planning rule only works if it survives the local marketplace reality. The safest approach is simple: build the forecast from the same core hierarchy everywhere, but let the event calendar, lead times and stock constraints change by market.
Practical checklist before placing the next PO
- Confirm the last uninterrupted sales window for the ASIN.
- Check whether the current period includes any stock-out distortion.
- Reconcile on-hand, reserved, inbound and unfulfillable stock.
- Check lead time from PO to sellable stock, not just to shipment receipt.
- Review upcoming promotions, price changes, PPC pushes and launches.
- Separate seasonal peaks from normal weekly demand.
- Review returns, cancellations and unusual spikes.
- Sense-check the Amazon forecast against your own records before buying.
- Escalate to human review if the forecast could be wrong enough to change the decision.
If you want the simplest working rule, use this: recent clean sales first, then supply position, then commercial context. Everything else is refinement.
Related reading and next steps
For a deeper planning stack, see how to forecast Amazon FBA inventory, which Amazon reports a brand should review every month, how to build an Amazon product contribution model and Amazon UK/EU launch operations checklist. If you are building a more disciplined replenishment process, start with the reporting hierarchy above, then turn it into one repeatable monthly view that operations, marketing and finance can all use. ([sellercentral.amazon.com](https://sellercentral.amazon.com/help/hub/reference/external/G561?locale=en-US&utm_source=openai))
Brian Vallois is the founder of Consultancy Worx, with 15 years of in-depth Amazon marketplace experience working with UK, European and US brands, including P&G. A consultant, investor and SaaS builder, he combines hands-on marketplace expertise with a focus on profitability, clear operating procedures and operational efficiency. His articles share practical lessons to help brand owners make better decisions and build businesses that are easier to manage.