StockSweep StockSweep | Help Center

Markdown Optimization Explained

How the recommended discount and expected recovery are calculated, why some SKUs show no recommendation, and how to read the confidence label.

Last updated: September 23, 2026

Contents
Show all sections (11) ▼

For each dead SKU, Markdown Optimization recommends the discount depth expected to recover the most money — not the deepest discount, and not a fixed default. Nothing on this page changes a price until you click Apply. This article explains where the recommendation comes from, so a number here reads as a calculation, not a guess.

Estimated time: 4 minutes

What you'll accomplish

  • Understand what "expected recovery" means and how a discount is chosen to maximize it
  • Know why some SKUs show no recommendation at all
  • Read the confidence label and the "Based on category" tag correctly
  • Know what "uplift vs. flat" is being compared against

Requirements

  • Plan: Pro or Enterprise.
  • Found at Analytics → Markdown Optimization.

How a recommendation is chosen

For each SKU, StockSweep tests a range of candidate discount depths and, for every one, estimates how likely the product is to sell at that price and how much money that would recover (price minus cost, when cost is known — price alone if it isn't). It picks the depth with the highest expected recovered value, not the deepest discount and not a flat percentage applied to everything.

Every candidate discount is checked against the exact same margin-floor and market-price rules that a manual discount you apply by hand goes through — a recommendation here can never suggest something your own safety settings would reject. If your margin floor or market bounds would clamp a deeper discount to a shallower one, the recommendation reflects the clamped depth, not the theoretical deepest one.

Where the sell-probability estimate comes from

The likelihood-of-sale-at-this-depth estimate blends two things:

  • How similar products in your store have responded to discounts at this age and category, historically.
  • This specific SKU's own price-cut history, if it has enough of one. A SKU with more of its own history is weighted more toward its own pattern; a SKU with little or no discount history relies mostly on the category pattern.

The "Based on category" tag you see next to a recommendation means the SKU didn't have enough of its own price-response history yet, so the estimate is drawn mainly from its product category's pattern rather than its own. That's not a lower-quality recommendation — it's the honest source for SKUs new to discounting.

Why some SKUs show no recommendation

A banner on the page explains that some active SKUs are "not currently actionable." That happens for one of a few reasons, and the page states which applies:

  • Insufficient history — not enough data yet to build a reliable response estimate for this SKU or its category.
  • Missing price or inventory data — the calculation needs a current retail price and quantity on hand to compute recovery. (Cost is separate: see the FAQ below — a missing cost downgrades the recommendation from profit-based to revenue-based, it doesn't block one.)
  • A binding safety floor — every discount depth tested would violate your margin floor or market bounds, so there is no discount left to recommend.

These SKUs stay visible with a reason shown, rather than being silently dropped from the list — a missing SKU you can't explain is worse than one flagged as blocked.

Reading "uplift vs. flat" and confidence

Uplift vs. flat compares the recommended discount's expected recovery against a flat baseline discount, so you can see how much better a tailored depth does than applying the same percentage to everything. Confidence (Medium/High/Low/Not enough data) reflects how much real data backs the estimate — it is a function of how much history is available, not how strongly the system "believes" the number. Low and "not enough data" confidence recommendations are shown, but treat them as a starting point to review rather than a number to apply blind.

FAQ

Does applying a recommendation here create a one-time discount or a schedule?

The page shows a single recommended discount per SKU; a staged schedule across multiple discount steps is a separate concept (see escalation chains and the response-curve "Sells in" estimate on this same page for expected time to sell).

New sales data (yours or your category's) shifts the underlying response estimate, and your inventory, cost, or margin settings can also change what's feasible. A changed recommendation reflects new information, not a random re-roll.

If I don't have cost data on a SKU, can I still get a recommendation?

Yes — without cost data, the calculation optimizes for recovered revenue instead of recovered profit. Adding real cost data is what upgrades a revenue-based recommendation into a profit-based one.