Unit 2/Lesson 1 of 3

Smart Replenishment — The Core Algorithm

Smart Replenishment is Tightly's forecasting engine: it ranks every SKU by urgency and revenue impact, shows Days of Stock, and generates buy/transfer recommendations with full reasoning.

SkillsForecasting algorithmsReplenishment logicAI product mechanics
+25 XP
✏️ On desktop — select text to ask Nobi, save notes, or add thoughts

How the composite scoring matrix works

Smart Replenishment doesn't just flag SKUs that are low on stock — it ranks them by a composite score of urgency and revenue impact simultaneously.

Why both dimensions matter:
- Urgency alone would surface every low-stock SKU equally. A slow-moving SKU with 5 days of stock is less important than a fast-moving SKU with 12 days of stock (if lead time is 15 days).
- Revenue impact alone would always prioritize your best-sellers. But if your best-seller has 45 days of stock, there's no urgency — prioritizing it wastes attention.

The composite score identifies the intersection: high-revenue SKUs that are approaching their reorder point. These are the items where acting now prevents the most damage.

Output: a ranked list, updated daily/weekly/monthly depending on your configured replenishment cycle, showing every SKU with its DoH, projected stockout date, recommended order quantity, and the supplier to order from.

Senteせんて

Having the initiative. A replenishment recommendation with urgency + revenue scoring gives the buyer sente — they act before the stockout forces a reactive crisis.

Inputs the algorithm consumes

The quality of replenishment recommendations depends entirely on the quality of inputs. Tightly pulls:

Live inventory — from Shopify and 3PL integrations (units on hand, per location)
Sales velocity — recent sales weighted by recency (last 7 days matters more than last 90 days)
Inbound POs — units already ordered and in transit (crucial for accurate DoH)
Supplier lead times — per supplier, ideally updated from actual delivery history
Safety stock policies — minimum coverage rules set per SKU or category
Budget constraints — total open-to-buy budget caps what gets recommended
Seasonal event flags — buyer-configured upcoming promotions or seasonal peaks

The critical one: inbound POs. If a PO for 500 units is already in transit and not captured in the system, Tightly would recommend ordering another 500 units unnecessarily. Accurate PO tracking prevents double-ordering.

Adaptive learning over time

Tightly's forecasting engine learns seasonality patterns from your brand's own sales history. First season: the algorithm uses a generic seasonality template. Second season onward: it adjusts based on what your brand's actuals looked like.

This matters because two brands in the same category can have wildly different demand shapes. A swimwear brand's peak might be March–May (northern hemisphere summer prep). A Scandinavian outdoor brand's peak might be October–November (pre-winter). Generic seasonality models get this wrong. Brand-specific learning gets it right.

As PM, the "learning flywheel" is a key differentiation narrative: the longer a customer uses Tightly, the better it gets for their specific business. This creates both retention (high switching cost as the model learns) and a competitive moat (a newer tool doesn't have your training history).

Ask Nobiexplain

Select any text in the lesson to ask a question. Press Esc to close.

1 / 3

Why does Tightly's composite scoring rank SKUs by BOTH urgency AND revenue impact rather than urgency alone?