Introduction

A Network view visualises how things are connected - whether it's people, processes or products. It helps you to see clusters, bottlenecks and connections that may not be obvious when looking at the raw data.

Omniscope provides a suite of Network views: Network, Tree, Sankey, Dependency Wheel and Sunburst. In this article, we'll explore each one of these views, and along the way we'll build a multi-tab report that brings them all together.

Demo Data

To bring the network views to life, we'll be using a fictional logistics scenario: Pawcels, an online pet food delivery company.

Pawcels manages everything from warehousing to doorstep delivery, ensuring their customers cats, dogs, fish and even stick insects never miss a meal.

For our demo, we've created four datasets that reflect the different parts of Pawcels' operations. 

Employees

This dataset contains the company's workforce. It includes each employees ID, Name, Role, Region, Depot and Reporting line. Approximately two-thirds of staff are drivers, who are responsible for completing deliveries, while dispatchers and managers oversee their work.

Deliveries

This dataset tracks the journey of every parcel. Each record includes the Depot, Delivery zone, Scheduled and Actual delivery times, Distance travelled and the Final status (e.g. on time, delayed, failed). 

Orders

This dataset captures what customers are buying on the Pawcels' website. Orders includes the Date, Zone, Product details, Pet type, Volume and Packaging. One order can contain multiple products.

Driver Handoffs

This dataset records handoffs - when one driver passes an order to another, often due to congestion, re-routing or zone boundaries. 

Network Types Explained

Before we dive in and start building our dashboard, it's useful to step back and look at the network view suite as a whole, and how different view types fit together.

Every network view in Omniscope shares the same network configuration options. In this section you tell Omniscope how to construct a network from your data. There are currently two main approaches: Grouped Networks and Linked Networks.

Grouped Network

A grouped network works by splitting the data step by step. You define a series of splits (groupings) and the network is constructed by breaking the data down at each level.

For example:

• Split 1 = Pet Type
• Split 2 = Product Name

The result is a tree-shaped network: the root represents all data, which is split into Dog, Cat etc. and each of these splits further into their respective products.

A grouped network is always tree like (there are no cycles) and the number of levels corresponds to the number of splits. Conceptually, it's similar to the drilldown feature in some views (e.g. Bar, Pie), except you can see the entire structure up front, rather than clicking to drilldown.

Grouped networks are highly flexible; they can be build from almost any dataset.

Linked Network

A linked network focuses on relationships between fields. You define two fields: Link Field 1 and Link Field 2. Omniscope will create a network where each node is a value in these fields, and displays the links between them.

For example:

• Link Field 1 = Superior Name
• Link Field 2 = Employee Name

This creates an organisation chart, where each superior is linked to their reporters.

In addition you can configure:

• The Link direction. This can be undirected or directed. In the example above we would choose a link direction of Superior Name -> Employee Name.
• The Node identifier. This determines how nodes in the network are uniquely identified. In the example above we would choose Employee Name as the node identifier. This is important if you want to size or colour nodes by a measure associated with another field (e.g. Salary).

Unlike grouped networks, linked networks can represent general graphs and can include cycles. This means they can be used to show trees, and more complex networks like collaborations or handoffs. Not all network views can display cyclic graphs, so you need to keep this in mind.

Shared Configuration Options

Beyond the network structure, many configuration options are shared across all network views, for example:

• Styling (node and link colours, sizes, shapes)
• Labels
• Tooltips

Switching between network views

When you switch from one network view to another, Omniscope will carry across your network settings and shared options automatically. This means you can switch your network to a Tree, a Sankey or a Dependency Wheel and see which works best for your data.

Network View

The Network view is a free-form graph of nodes and links. It's useful for conveying structure in your data. With large datasets it can become visually dense. particularly when there are lots of interconnecting dependencies between nodes.

Example 1: Employees (Grouped Network)

Lets start by building a network view with the Employees dataset.

1. Add a new Network view, ensure the Employees dataset is selected.
2. Open the Network options and select Grouped Network.
3. Add the following splits: Region, Depot, Employee Name.
4. In the Node options, select size and add a Record Count measure.

The resulting network breaks down all records (employees) across these dimensions. We can immediately see, for example that the South region contains the largest number of employees.

By default the layout is set to Spring (a force-directed layout). This simulates physical forces, causing nodes to spread out and settling into clusters. You can Enable or disable the animation (to watch the nodes fall into position). You can also adjust the number of iterations. Higher values may produce a more stable network, but increase the positioning time.

Other layout types include:

• Balloon - nodes radiate outward from a central root.
• Tree - arranged in a traditional tree structure.
• Circle - nodes places evenly around a circle, with optional ordering.
• Grid - rows and columns with optional ordering.
• Spiral - nodes follow a spiral path.
• Manual - maps node position using X and Y values derived from the dataset. This is useful for pre-calculated layouts, or approximating real-world coordinates like depot locations.

Example 2: Employees (Linked Network)

Using the same dataset, we can switch to a linked network.

1. In the Network Options, select 'Linked network'.
2. Linked Field 1 = Superior name
3. Linked Field 2 = Employee name
4. Direction = Field 1 -> Field 2
5. Node identifier key = Field 2.

The configuration produces an organisational chart, showing the company's reporting structure. To enhance, open the node options and size the nodes by Salary. 

Senior roles now appear more prominent. We can see Finley Robinson (CEO) at the root of the tree, with regional managers (e.g. Alex Jackson) and their teams below.

Example 3: Driver Handoffs (Linked Network)

Finally, lets use the Driver handoffs dataset to configure another linked network.

1. Switch the dataset to the Driver handoffs dataset, and 
2. Linked Field 1 = From Driver Name.
3. Linked Field 2 = To Driver Name.
4. Direction = Field 1 -> Field 2.
5. Node identifier key = Field 2.

This produces a graph rather than a tree. To reveal patterns, open the node options and colour by 'To driver depot'.

The result shows clusters of drivers working within the same depot, with occasional cross depot handoffs. Most transfers are local, but you can see some bridges where drivers assist across depot boundaries.

Tree View

The Tree View is used for displaying hierarchical data. Although the Network View can also show trees, this view is dedicated to presenting them in a clean, structured format.

We won't go into detail on the underlying mechanics here, as that's covered in this article - instead we'll focus on applying it to our data.

Example: Employees (Linked Network)

1. Add a Tree View and select the Employees dataset.
2. Choose Linked Network.
3. Link Field 1 = Superior Name.
4. Link Field 2 = Employee Name.

This produces the same organisation hierarchy we saw earlier in the Network view, but displayed in a clear tree structure. 

You can flip the layout to display the tree vertically or horizontally, depending on what fits best with your dashboard design.

Sankey View

The Sankey view is used to show flows - how data moves from one category to another. Each link has a width proportional to its weight, so you can immediately see where the largest volumes flow. Sankey diagrams are often used for showing things like energy flows, or in this case delivery performance.

Example: Deliveries

We'll use the Deliveries dataset to show how orders flow from Depot -> Delivery Zone -> Status.

If we build this as a grouped network, each zone would get its own repeated set of Status nodes. Instead we'll create a linked network, where  the three Status nodes (On Time, Delayed, Failed) are shared across all zones.

To do this, we prepare the data in the workflow:

1. Aggregate the data by Depot and Delivery Zone. Rename the fields to 'From' and 'To' and add a record Count measure.
2. Aggregate the data by Delivery Zone and Status. Rename the fields to 'From' and 'To' and again, add a record count measure.
3. Append the two datasets together and connect this dataset to the report.

Now switch to the report:

1. Add a Sankey View and select the Deliveries (aggregated) dataset.
2. Switch to Linked Network.
3. Link Field 1 = From
4. Link Field 2 = To
5. Under the link options, set the Count field as the link weight.
6. Set node and link colour to the From field.

The result is a Sankey diagram showing delivery volumes flowing from depot, through zones and into status categories. For example, hovering over the link Cardiff -> Cardiff Bay shows 26 deliveries, and following the path further reveals that 21 were on time.

This makes it easy to compare both the scale of each depot and the performance outcome across zones.

Dependency Wheel

The Dependency Wheel view can be used to show pairwise relationships between categories. Each value is represented as a node around the outside of the wheel, with curved links connecting items. The thickness of each link is proportional to the weight value.

Example: Order Items

We'll use the Order Items dataset to show which products are most often bought together. For example, if a customer orders Clawlicious Salmon Bites, what other products typically go in the same basket.

The raw Order items dataset looks like this (simplified):

Order ID     Product Name

OD001        Clawlicious Salmon Bites

OD001        Nibble Nuggets

OD002        BarkBites Supreme

OD002        Woof & Ready Chunks

As we can see, each order can contain multiple rows - one per product. To build a dependency wheel, we need to transform this into a linked dataset, with three fields:

1. Product 1
2. Product 2
3. Count (how many times they appeared in the same order)

Preparing the data

You could produce this data manually in the workflow, for example:

1. Group by Order ID, collecting all products in each order.
2. Generate pairs of products within each order.
3. Aggregate across all orders to count how often each product pair occurs.

Alternatively, in this example we are going to use the AI block to do the heavy lifting:

1. Add an AI block.
2. Type in a prompt, such as: "This dataset contains a list of orders. Each order has a list of products. I want to create a dataset showing associations between products — which products are bought together. The dataset should contain three fields: Product 1, Product 2, and Count. ".
3. Run the block. The AI will generate some Python code to produce the associations dataset automatically.

Visualising the associations

1. Connect the new Product Associations dataset to your report.
2. Add a Dependency Wheel view and select this dataset.
3. Switch to Linked network.
4. Link Field 1 = Product 1
5. Link Field 2 = Product 2
6. Weight = Count.
7. Colour links by Count.

You'll now see a wheel of products, with links showing how often they co-occur. For example, you can see a thicker, darker connection between Clawlicious Salmon Bites and Purrfect Pouches; hovering over the link shows they were ordered together 8 times.

This type of view is valuable for making marketing decisions (e.g. bundle promotions) and also for operational efficiency (e.g. storing items frequently bought together closer in the warehouse).

Sunburst View

The Sunburst view is like a multi-level Pie view. Each ring represents a level in the hierarchy, with segments sized by a chosen measure (such as volume or count). Unlike other network views, the Sunburst doesn't display links - it focuses on showing proportions within each level of the hierarchy.

Example 1: Orders

We'll start by exploring the Orders dataset.

1. Add a Sunburst View and select the Orders dataset.
2. Choose Grouped network and configure the following splits: Pet Type, Product Name.
3. Select Volume (KG) as the measure.
4. Colour nodes by Pet Type.

The resulting Sunburst shows the order breakdown by weight. At the first level we can see that Dog food accounts for the largest share. At the second level, we drill down into products - for example Canine Crunchies XL contributes the most within dog food.

Because each level is relative to it's parent, we can quickly spot category leaders. This can help with stock management (knowing which items to keep in higher supply), warehouse planning (organising our shelves with the most frequently ordered products), marketing (highlighting bestsellers to customers).

Example 2: Employees

The Sunburst view can also visualise tree-shaped linked networks (but without links)

1. Add another Sunburst View and select the Employees dataset.
2. Choose Linked Network.
3. Field 1 = Superior Name
4. Field 2 = Employee Name

The view now displays the company hierarchy as concentric rings: the CEO at the centre, regional managers in the next ring and their teams around the outside. It's a more compact way to view the organisational structure without the complexity of a full network layout.

Summary

The Network View suite in Omniscope provides a versatile set of tools for visualising relationships, hierarchies and flows. Each view has it's own strengths:

• Network View for flexible force-directed exploration of clusters and graphs.
• Tree view for clean, structured hierarchies.
• Sankey for directional flows and proportional volumes.
• Dependency Wheel for pairwise associations.
• Sunburst for hierarchical proportions in a compact radial form.

Choosing the right view depends on the structure of your data. In some cases you'll need to prepare the data before building a linked network, e.g. for creating an associations table for the dependency wheel, or restructuring the deliveries dataset for the Sankey. But once the network is configured you can easily switch between views to see your data from different perspectives, reusing the same configuration and style options.

Looking ahead, we have a number of ideas for enhancing the Network view suite further, for example:

• Additional layouts in the Network view.
• Network types that automatically generate co-occurrence or flow datasets, reducing the need for manual preparation.

If you have any ideas, or feedback let us know, we'd love to hear from you!