Amazon SKU rationalization is the process of deciding which products deserve continued investment, which need a specific repair and which should leave the active assortment. It is not a rule that tells you to delete every low-selling product.
I review the product decision before the score. A useful assortment review tells the team what to do next, who owns it and when to reconsider.
A useful review connects demand, contribution, inventory and the product's role in the range. A seasonal item, an unavailable bestseller and a genuinely unwanted product can all show weak recent sales. They need different decisions.
The output should be an operating plan: a decision, an owner, a spending boundary and a review date for each product that needs attention.
Define the unit of analysis first
A parent ASIN, a child ASIN and a merchant SKU are not interchangeable. A parent-level sales total may hide an unprofitable size or color. A SKU may represent a fulfillment or operational distinction that does not require a separate marketing strategy.
Create a mapping that connects the commercial product family to its child ASINs, merchant SKUs and marketplaces. Use family-level views for positioning and range decisions, then inspect variation-level economics and availability before acting.
This matters when advertising reports credit sales across products or variations. Document how ad costs are assigned. If you cannot allocate them reliably, label the estimate instead of presenting a precise but misleading SKU profit figure.
Build a review table you can trust
Start with these fields rather than a complicated score:
| Field | Decision it supports |
|---|---|
| Product status and last verification date | Is this active, launching, seasonal or discontinued? |
| Units and net sales over comparable periods | Is demand changing? |
| Contribution before and after allocated advertising | Is selling the product economically useful? |
| Sellable stock, inbound stock and lead time | Can demand be served? |
| Stock age and carrying costs | What does waiting cost? |
| Availability and material offer changes | Is recent performance representative? |
| Returns and recurring product complaints | Is the product creating a customer problem? |
| Evidence quality | How confident should the decision be? |
Reconcile dates, currencies and marketplaces. Keep unavailable data visibly missing. A blank cost is not a zero cost.
Before creating product actions, confirm the discontinued list with its actual owner. An outdated status field can send a team to improve products it no longer intends to replenish.
Read recent demand in context
Use a recent window to detect change and a longer or comparable seasonal window to understand context. The appropriate periods depend on order volume and the buying cycle; there is no universally correct 30-day cutoff.
Mark promotions, stockouts, price changes and major traffic changes. For event-driven products, compare an event period with a relevant event baseline and ordinary trading with ordinary trading. Do not silently mix the two.
For a simple stock indicator:
Days of cover = sellable units on hand ÷ expected daily unit demand.
This is a planning estimate. Past sales during a stockout are not a clean forecast of unrestricted demand. Zero recorded sales should trigger investigation, not an automatic division or an unquestioned “infinite inventory” conclusion.
Inbound units need their own date and confidence level. A purchase order is not the same as inventory available to ship today.
Use decisions with explicit entry conditions
The following is a proposed review framework, not an Amazon classification system:
| Decision | When it may fit | Required next step |
|---|---|---|
| Invest | Credible demand, positive economics and feasible replenishment | Define the incremental budget and stock boundary |
| Repair | A specific, testable constraint explains weak performance | Assign one intervention and a review condition |
| Maintain | The product earns its role but does not justify expansion | Keep sensible support and review periodically |
| Exit | No credible future role or an approved discontinuation decision | Stop replenishment and assess stock recovery options |
| Investigate | Critical inputs are missing or contradictory | Resolve the data gap before committing spend |
Avoid a weighted score that lets strong revenue cancel out a serious quality issue or missing costs. Some conditions should be gates. A product with an unresolved safety or compliance concern does not become an investment candidate because its sales are high.
Example: similar sales, different actions
Consider four hypothetical products. Each illustrates a different commercial decision.
Product A sells slowly but earns strong contribution. It is a specialist variation that completes the range, with modest carrying cost. Maintaining it may be more sensible than deleting it to simplify a report.
Product B shows falling sales after repeated stockouts. Increasing advertising does not solve the immediate problem. The next decision belongs to replenishment and offer availability.
Product C has excess stock and repeated complaints about a product feature. A better headline is unlikely to address the cause. Investigate the product issue and whether a continued assortment role is defensible.
Product D has weak economics because the current promotion is too deep. Model ordinary pricing, realistic demand and variable costs before deciding that the product itself is unviable.
The point is not to defend every SKU. It is to make each intervention match the evidence.
Turn the analysis into a team workflow
A catalog review fails when “fix” becomes a permanent parking lot. Every repair decision needs a testable explanation and someone accountable for the next action.
A useful action record contains:
- Product and marketplace.
- Observed problem and supporting evidence.
- Hypothesis to test.
- Action owner and dependencies.
- Budget or effort limit.
- Success, stop and reconsideration conditions.
- Review date and resulting decision.
For example: “The featured variation is unavailable while ads still promote it. The operations owner will confirm replenishment; the advertising owner will align the promoted variation; review after availability and reporting stabilize.” That is more actionable than “improve conversion.”
The published product, stock and returns workflow case offers related project context. It shows how product status, stock and customer feedback can be organized into a working process.
Keep portfolio decisions separate from liquidation decisions
Deciding not to reorder answers a future investment question. Deciding what to do with inventory already owned answers a recovery question. The best answer can differ.
A discontinued product may still be worth selling through at a controlled contribution. Another may lose more value through storage and delay than it can recover through normal sales. Compare the available recovery routes before implementing the exit.
For ongoing products, coordinate stock decisions with inventory-aware Amazon PPC. For a broader diagnostic, review Amazon growth and inventory analytics.
What to bring to a catalog review
Start with a product list, verified lifecycle status, recent and comparative sales, variable costs, stock and replenishment assumptions. Note where allocation or data quality is uncertain.
Send a project brief with your marketplace, approximate assortment size and the decision your team is struggling to make. The first useful question is not “How many SKUs should we cut?” It is “Where is the next unit of budget or attention most defensible?”
Continue the analysis
- Slow-Moving Amazon Inventory: Discount, Hold or Liquidate?
- Amazon External Traffic: How to Plan a Profitable Test
- Inventory-aware Amazon PPC
Common questions
Should every unprofitable SKU be discontinued?
Not automatically. First establish whether the loss is persistent, whether costs are allocated correctly and whether a realistic repair or strategic role exists. Do not use speculative future value to excuse indefinite losses.
How often should a portfolio be reviewed?
Review urgent exceptions when they arise and use a regular portfolio cadence suited to your replenishment and demand cycles. Low-volume products may require longer observation periods than high-volume products.
