Tutorial: Supply Chain Risk Analysis #

This tutorial demonstrates the workflow of SemSpect using the Northwind dataset.

The Business Scenario Northwind Traders needs to identify vulnerabilities in the trade network by answering a critical question:

“Which VIP US Customers are at risk if the UK Supply Chain fails, and which product lines will be impacted?”

We will model the supply chain, filter for high-volume partners, and isolate the “at-risk” product mix.

1. Seed the Canvas #

Every exploration requires a starting point. We begin with the central entity: the Product.

1. Locate label Product in the sidebar tree.
2. Drag it to the center of the canvas.
3. Result: A group appears, representing the entire inventory (e.g., 77 Products).

2. Map the Flow #

Next, we expand the graph to connect Suppliers and Customers.

Preparation #

To simplify the expansion menus, hide generic labels that are not relevant to this analysis.

• Right-click the label Agent in the sidebar.
• Select Deactivate label.

Expansion Steps #

We extend the graph from the central Product node.

1. Add Suppliers:

   • Select the Product group.
   • Expand via label Supplier → hasSupplier.

2. Add Customers (via Orders):

   • Select the Product group again.
   • Expand via label Order ← includesProduct.
   • From the new Order group, expand via label Customer → hasCustomer.

3. Filter Key Players #

The graph currently visualizes all connections. We must now isolate the high-value targets (VIPs) and the specific risk region (UK).

Identify VIP US Customers #

We define “VIP” as customers with the highest volume of incoming orders.

1. Select the Customer group to open the Node Table.
2. Filter by Country: Set country to USA.
3. Identify VIPs: Click the Predecessors column header to Sort Descending. This ranks customers by their incoming connection count (number of Orders).

4. Select & Restrict: Check the boxes for the top 5 customers and click Restrict to selection.

Isolate UK Suppliers #

1. Select the Supplier group to open the table.
2. Filter Country to UK.
3. Click Apply filters.

The Disconnected State

At this stage, you have filtered the ends of the chain, but the center (Products) still shows the total inventory. The graph displays the topology, but the filter logic hasn’t propagated to the middle yet.

4. Propagate Risk (Backpropagation) #

We now need to identify the intersection: Which products are both sold by UK Suppliers and bought by VIP US Customers?

We use Backpropagation to push the filters from the outer nodes inward to the Product node.

1. From Supplier: Right-click the connection between Supplier and Product. Select Restrict left group (this propagates the UK filter to the Product list).
2. From Customer: Repeat the restriction logic from the Customer side:
   • Order ─ hasCustomer → Customer
   • Product ← includesProduct ─ Order

Result: The Product group count drops. These remaining items are the specific products at risk.

5. Analyze Impact #

Now that we have the “At-Risk” product set, we analyze it to understand the business impact.

1. Select the filtered Product group to open the Histogram Panel.
2. Expand the Product Labels (e.g., Beverages, Confections).
3. Insight: Identify which bar is largest (e.g., “Confections”). This reveals that a UK supply failure will primarily impact the dessert product line.

6. Operationalize Findings #

Finally, we convert this visual insight into an actionable list for the procurement team.

1. Drill Down:

   • In the Histogram, right-click the “Confections” bar.
   • Select Restrict group to label. This discards the other product types, leaving only the high-risk Confections.

2. Trace the Branch:

   • Open the Node Table for this group.
   • Right-click a specific high-value item (e.g., “Teatime Chocolate Biscuits”).
   • Select Flag all connected nodes on the branch.

• Visual Output: SemSpect highlights the exact path of risk: The specific UK Supplier → The specific Biscuit → The specific VIP US Customers buying it.

Summary & Next Steps #

You have successfully modeled a complex dependency network, filtered it by geo-location and volume, and identified a semantic risk cluster.

• Share: Save the Exploration to share this view with colleagues.
• Monitor: Create a Custom label called “High-Risk Confections” to permanently track this segment.
• Report: Export the CSV of the flagged branch for the procurement team.