Distribution network design requires a clear understanding of customer groups from a spatial viewpoint – with group-specific preferences and needs as well as other relevant characteristics. This is why the simple SCDA geocoding and heat mapping app was used by a small but quickly growing e-shop in the US. Applying the SCDA heat mapper was the first step towards developing a distribution strategy and therewith associated distribution network design. Main objectives were improved customer satisfaction, i.e. faster sales growth as a result of improved distribution efficiency (faster delivery, faster returns, and cheaper delivery).

The app applied in this project was the SCDA geocoding and heat mapping app. It is available for download via below link.

• Link: Geocoder & heat mapping app

Below video displays a quick demonstration of the SCDA geocoding and heat mapping web application used by the client for this project.

First, customer clustering criteria were defined and applied for grouping customers

Customers were grouped into five (5) clusters:

• Large retailers, regulars: Customers with warehousing capacity and longer purchasing cycles. Repeating customers
• Large retailers, irregulars: Large retailers that are either one-time customers, or have very irregular purchasing cycles and volumes
• Small retailers: Retailers with no or only very limited inventory
• Drop shipping customers: Private consumers buying from unbranded e-commerce branch, i.e. mostly consumers receiving product at their home address. Shipping directly from the manufacturer, with production upon receiving customer order
• Branded customers: Customers receiving products at their home address, purchased through the main and branded web shop, with shipping from stock at central warehouse (not from manufacturer)

The vendor had a good existing understanding of relevant customer clusters. Generic clustering criteria for an analysis of this kind are however e.g. the following:

• Ordering lot sizes, e.g. minimum ordering quantities
• Consignment stock vs. non-consignment
• Inventory vs no inventory
• Expected delivery lead time
• Process for order placement (e.g. through web shop, physically in a store, via sales agent, etc.)
• Sales volume
• Pricing power
• Breadth of the claimed product range
• Inbound transport characteristics of claimed product range
• Ordering frequency
• Ordering reliability
• Packaging requirements
• etc.

Second, historic transaction data was geocoded and visualized with the SCDA heat mapper

Using the SCDA geocoding and heat mapping app, a simple application available in the SCDA shop, customer locations were geocoded and historic shipping data was visualized in the form of distribution intensity heat maps. The relevant KPI may be any relevant transactional metric. In this case, sales revenue in USD.

While above results were not a total surprise to the e-commerce vendor this analysis is still often important and necessary for obtaining a confirmed overview of spatial customer distribution by cluster. Network design or optimization criteria can then be defined, based on the confirmed overview obtaining throughput customer clustering and customer cluster visualization.

Third, a distribution network was designed with cost and network optimization models

With a clear overview of customer needs and spatial customer distribution efforts a network design and distribution strategy was drafted. A new warehouse was installed as a distribution hub, with beneficial proximity to both relevant customer groups as well as inbound container terminals for supply from overseas. A cost model, taking into account variable transportation costs, variable warehousing costs, and fixed warehousing costs, as well as costs for alternative 4PL logistics services, was developed and applied – confirming one out of several warehouse facility location candidates.

While not applicable in this project such calculations and their results can be visualized in app, too. Furthermore, relevant input parameters such as e.g. potential location candidates and capacities, cost and price data, freight zones and tariffs, etc., can also be parametrized and made available to the user through an interface. This can prove helpful when communicating results, e.g.  for acceptance and change management for final implementation.

Concluding remarks related to customer clustering and distribution strategy drafting

This article summarized an exemplary application of the simple SCDA heat mapper, a simple tool for geocoding customer location data and shipping, sales, or other relevant spatial transaction data. This tool was used as the first step in a network design project, with the main purpose of its application being exploratory data analysis of spatial customer transaction data. The main result of such an analysis is an improved overview of spatial customer distribution, from both a sales and logistics point of view.

Here are some other SCDA articles and products with related content:

• Link: Prescriptive analytics for network design
• Link: Distribution network visualization (Python)
• Link: Simulation for yogurt supply chain optimization

Linnart Felkl

Data scientist focusing on simulation, optimization and modeling in R, SQL, VBA and Python