Product Health: Decide What to Fix First

Marketplace portfolios generate more warnings than teams can act on. Stockouts, inactive offers, Buy Box losses, weak margin, and listing issues all appear urgent when shown in red.

Prioritization needs three dimensions: commercial impact, severity of the broken state, and realistic recoverability.

1. Create a shared health language

Define mutually understandable states. “Unhealthy” is too vague. Use operational categories such as out of stock, inactive, lost box, unprofitable, weak listing, or data incomplete. A product can have more than one state, but one should lead the action queue.

Keep health separate from performance. A product can be healthy but small, or high-revenue but structurally unhealthy.

2. Apply a severity order

Start with states that block all sales:

1. Out of stock or inactive.
2. Lost selling position or offer eligibility.
3. Negative unit economics.
4. Conversion and listing-quality constraints.
5. Growth opportunities on otherwise healthy products.

This order is a default, not a replacement for impact. A tiny inactive product may matter less than a major product losing margin rapidly.

3. Add commercial impact

Estimate revenue or contribution at risk using a recent stable baseline. For stockouts, include expected lost units and likely replenishment duration. For Buy Box loss, compare the previous share or sales rate. For unprofitable products, use contribution loss rather than revenue.

Rank absolute euro impact alongside percentage severity. This directs limited operator time toward material problems.

4. Score recoverability

Some issues are fixable today; others depend on suppliers, marketplace review, or missing data. Add time to recover, confidence in the diagnosis, and action ownership.

5. Work one queue across channels

Avoid separate disconnected lists for stock, ads, profit, and content. A product-level view should show which channel is affected and link the operator to the place where the action occurs.

Global diagnosis is useful for prioritization; connection-level context is required for execution. Preserve that relationship so a team does not “fix” the wrong account or country.

6. Prevent false recovery

A product is not healthy because the badge disappeared. Confirm the commercial outcome. Restocked products should recover availability and sales; repriced products should regain competitiveness without destroying margin; listing fixes should improve conversion after enough traffic.

Keep the original issue and action history. Recurring stockouts or repeated Buy Box losses reveal a process problem.

7. Practical operating rhythm

Review critical blockers daily, material margin and competitiveness issues weekly, and lower-severity optimization opportunities monthly. Use service-level expectations: who owns the issue and when the next update is due.

Limit work in progress. Closing five high-impact issues is better than opening fifty low-confidence investigations.

8. Worked example: turning 120 warnings into five actions

Suppose a portfolio health view flags 120 product-channel combinations. Thirty are out of stock, twenty are inactive, fifteen lost the Buy Box, twenty-five are unprofitable, and thirty have weak listing grades. Treating each warning equally overwhelms the team.

First, aggregate by product and calculate recent contribution or revenue at risk. The thirty stockouts may reduce to four products explaining most expected loss. Some inactive offers may be old assortment with no strategic value. Several unprofitable flags may share one missing COGS source rather than separate commercial problems.

Create a top-five queue: replenish the two material stockouts, resolve one high-value inactive offer, investigate one Buy Box loss, and complete cost data for a product whose profit state is uncertain. Listing improvements move to the next queue unless they block a major campaign or explain a material conversion decline.

At the next review, confirm outcomes rather than closing tasks. The replenished product must become available; the inactive offer must sell or return to a healthy eligible state; the Buy Box action must preserve margin. If a warning recurs, promote it from product task to process problem.

Queue fields

• Product, channel, and account.
• Leading health state and secondary states.
• Revenue or contribution at risk.
• Confidence and recoverability.
• Owner, next action, due date, and dependency.
• Resolution evidence and recurrence count.

9. Practical checklist

• Health states have explicit definitions.
• Blocking states are separated from optimization opportunities.
• Queue uses contribution or revenue at risk.
• Recoverability and time to act affect priority.
• Global issues retain channel/account context.
• Each item has one owner and next review time.
• Resolution requires outcome confirmation.
• Recurring issues become process-level actions.
• Missing data is shown as its own health problem.

Product health becomes useful when it directs action. The winning interface is not the one with the most signals; it is the one that helps a team decide what to fix now, next, and not yet.

Calibrate the queue over time

Compare predicted impact with actual recovery. If high-priority items repeatedly produce little value, revisit the impact estimate or severity model. If lower-ranked problems become urgent before review, adjust cadence or leading indicators. Prioritization should learn from outcomes.

Give each team a work-in-progress limit. Operators should finish or explicitly hand off current issues before opening more investigations. Use blocked states for supplier, marketplace, or finance dependencies so work does not disappear inside personal notes.

Review the long tail monthly. Some products should be retired, cleared, or deliberately deprioritized rather than repeatedly repaired. Document that choice so the same warnings do not re-enter the queue every day. A health system is credible when it supports both action and intentional inaction with visible reasons.

Keep thresholds visible to operators and review them when economics or marketplace conditions change. A severity rule that made sense under old fees, margins, or delivery promises can misrank today’s work. Stable definitions matter, but unexamined definitions create false precision.