---
name: supplier-selection
description: How to rank and pick a supplier for a SKU. Load this whenever a task involves choosing a supplier, comparing quotes, or creating a purchase order.
---
<!-- Copyright 2026 Anthropic PBC -->
<!-- SPDX-License-Identifier: Apache-2.0 -->


# Supplier Selection

Ranking suppliers is arithmetic, not judgment. **Compute it in Python via code execution — do not reason about it in prose.**

## Method

For a given SKU:

1. Read `/mnt/user/data/supplier_catalog.csv` and filter to rows matching the SKU. This gives you `(supplier_id, unit_price, min_order_qty)` for each candidate.
2. Join with `/mnt/user/data/suppliers.csv` on `supplier_id` to get `lead_time_days` and `reliability`.
3. Normalize price and lead time across the candidates (min-max to [0,1], where 0 is best). Reliability is already on [0,1] where higher is better.
4. Score each candidate:
   ```
   score = 0.5 × (1 − norm_price) + 0.3 × (1 − norm_lead_time) + 0.2 × reliability
   ```
5. Pick the highest score. **Tie-breaks:** lowest `unit_price`, then lowest `lead_time_days`, then alphabetical `supplier_id`.

## Do this in code

Write and run a short Python script — don't call a tool per supplier, and don't compare quotes by describing them. Example shape:

```python
import csv
sku = "SKU-0057"
catalog = [r for r in csv.DictReader(open("/mnt/user/data/supplier_catalog.csv")) if r["sku"] == sku]
suppliers = {r["supplier_id"]: r for r in csv.DictReader(open("/mnt/user/data/suppliers.csv"))}
# join, normalize, score, sort — print the winner as JSON
```

## Supplier-specific overrides

These quirks are **not in the catalog data**. Apply them after computing the
score; if one changes your pick, say so in the rationale.

| Supplier | Override |
|---|---|
| SUP-01 Cascade Distribution | Orders >500 units need 48h notice or they auto-split into two shipments — add to lead-time math for large POs. |
| SUP-02 Alpine Wholesale | Closed Dec 20 – Jan 3. POs in that window aren't acknowledged until Jan 4. Land holiday replenishment before Dec 15. |
| SUP-03 Backcountry Supply Co | Two short-ships on Tents & Shelter this year. Prefer an alternate for that category if lead is comparable. |
| SUP-04 Sierra Outfitters | Unlisted 3% price break at 250+ units. If recommended qty is 200–249, often worth rounding up. |
| SUP-05 Granite Gear Partners | West-coast DC only. Add 2–3 days to catalog lead time for WH-EAST deliveries. |
| SUP-07 Ridgecrest Imports | Import-only; lead times are port-congestion sensitive. Don't rely on stated lead for an urgent order. |
| SUP-09 Summit Source | MOQ is enforced strictly — they reject (not round up) below-MOQ POs. |
| SUP-12 Trailhead Mercantile | New on roster. Treat reliability as one notch lower than stated until 6 months of history. |

## Warehouse lead-time adjustments

- **WH-WEST (Reno)**: most suppliers ship from east-coast DCs — add **+2 days**
  to catalog lead time as a rule of thumb unless a supplier note says otherwise.
- **WH-EAST (Carlisle)**: two-shift receiving dock; best destination for an
  expedited PO that needs same-day inbound scheduling.
- **WH-CENTRAL (Kansas City)**: overflow / transfer hub. If you're sourcing an
  inter-warehouse transfer rather than a PO, WH-CENTRAL is almost always the
  origin.

## Expedite override

If the reorder-policy skill flagged **expedite**, ignore the score and pick the supplier with the lowest `lead_time_days` whose `min_order_qty` ≤ your target order qty.

## Output

Return `{"supplier_id": "SUP-03", "unit_price": 21.40, "lead_time_days": 7, "score": 0.81}` and use that `supplier_id` in the PO.