Process Automation - Noverstock

In today’s fast-paced commerce landscape, businesses face increasing pressure to process orders quickly, reduce costs, and maintain accuracy. The challenge is even greater for companies handling massive volumes across multiple platforms. That’s where we come in.

Our system is more than just an automation tool—it’s a highly customizable operation-driven back-office solution designed to process large-scale orders while reflecting the complexities of real-world workflows. Let’s dive into the unique principles and innovations that set us apart.

Breaking Away from the Norm: Operation-Driven Automation

Most systems in the industry rely on event-driven workflows, where actions are triggered in response to specific events, often leading to rigidity and inefficiency. We took a different approach: operation-driven automation.

What Does Operation-Driven Mean?

Instead of reacting to events, our system operates based on defined roles and responsibilities that mimic how real humans manage order workflows. For example:

A Warehouse Picker focuses exclusively choosing which warehouse will fulfill an order
The Shipping Service Selector determines the optimal courier and service, factoring in costs, risks, destinations, weights, etc…

This separation of roles ensures each decision tree is highly specialized, scalable, and adaptable to unique business needs.

Decision Trees: The Backbone of Our System

At the heart of our system are decision trees—custom logic frameworks designed to automate decisions for specific operational goals.

How They Work

Each tree is tied to a defined goal and can emit specific actions. These trees are:

Rule-Based Yet Dynamic: While built on clear rules, they adapt to the internal state of orders.
Focused on Precision: Trees reflect business roles, preventing overlaps or conflicts in decision-making.
Future-Ready: We’re building an AI-powered tool to help users create and manage these trees, democratizing access to advanced automation.

Proprietary Language for Decision Trees

We developed our own decision tree language to:

Enable easy declaration of rules and logic.
Test and validate decision tree behavior in real time, ensuring they work as intended.
Optimize performance with a custom-built compiler that processes decisions quickly, even under heavy workloads.

The Lifecycle of an Order

Our system ensures every order is processed with precision from the moment it enters the pipeline. Here’s how:

Fetching and Mapping:
- Orders are pulled from external platforms by the Fetcher.
- The Data Mapper standardizes and cleans data, ensuring it’s ready for processing.
Validation:
- Orders undergo thorough checks for state, source authenticity, and data integrity.
- This step eliminates errors before orders progress further.
Pre-Creation Decision Trees:
- Order Processing Ranker: Prioritizes handling based on criteria like urgency or customer type.
- Order Status Changer: Updates the order’s status based on business rules.
- Default Value Resolver: Fills in missing data fields for seamless downstream processing.
Order Creation:
- Once validated, the order is created in the system and stored securely.
Post-Creation Decision Trees:
- Assign Flags and Tags: Identifies fraud risks or applies custom tags to control order flow.
Automatic Actions Queue:
- Orders enter a loop of decision trees for further optimization, handling tasks like shipping service selection, warehouse operations, and profit calculations.
User Queue:
- Fully processed orders are ready for user-facing tasks like fulfillment or reporting.

Deep Dive into the Automatic Actions Queue

At the heart of our system’s operational brilliance lies the Automatic Actions Queue. This queue is where the system truly shines, running a series of highly specialized decision trees to optimize every aspect of an order.

How It Works

When an order enters the Automatic Actions Queue, it undergoes a sequence of decision trees, each designed to address a specific part of the order lifecycle. What sets this apart is its iterative nature: if any tree makes changes to the order, the queue reevaluates the updated order to ensure all conditions are met.

Key Decision Trees in the Queue

Warehouse Picker
- Determines which warehouse should handle the order, optimizing for proximity, stock levels, and operational efficiency.
- Ensures minimal transit times for items, improving overall delivery speed.
Product Matcher
- Matches items in the order to products in the internal nomenclature.
- This ensures consistency, even when data from external platforms is incomplete or inconsistent.
Assign Package
- Evaluates the dimensions, weight, and nature of the items in the order to suggest the most cost-effective and suitable packaging.
Assign Order Attributes
- Determines critical metadata for the order, such as:
  - HS Codes: Harmonized System codes for customs declarations.
  - ADR Flags: Identifies whether the order contains hazardous materials requiring special handling.
  - Content Descriptions: For detailed shipping documentation.
Shipping Service Selection
- - Destination country and associated risks.
    - Price and weight constraints.
    - Service options like insurance, delivery times, and courier reliability.
    - Handles complex rules, such as prioritizing cost savings for low-margin goods or ensuring insured shipping for high-value items.
Assign Forecasted Shipping Cost
- Calculates an estimated shipping cost based on the selected service and package details.
- Provides valuable insights for profit calculations later in the process.
Warehouse Quantity Reservation
- Decides whether to reserve stock for the order, ensuring high-demand items are appropriately allocated.
Change Order Status
- Updates the order’s status if specific criteria are met, such as completing fulfillment steps or detecting anomalies during processing.
Reduce Order Quantities
- Evaluates whether the quantities in the order should be adjusted, often due to stock shortages or customer preferences.
Project and Seller Resolver
- Links the order to the correct internal project or seller, especially useful for businesses operating with multiple sales channels or regions.
Profit Calculator
- Uses heuristic formulas to determine the most accurate profit margin for the order.
- Factors in costs such as product price, shipping, warehouse handling, and operational expenses.
ProcessREDUCE_ITEM_UNITS
- Addresses scenarios where items in the order need quantity adjustments, ensuring that operational priorities are respected.

Scalability and Reliability: Built for Growth

Our system isn’t just intelligent—it’s built to handle growth and high demands. Here’s how we ensure reliability:

Horizontally Scalable Executors: Decision tree executors can scale horizontally to process thousands of orders per second.
Message Queues: Efficiently manage communication between components, ensuring smooth workflows.
Proprietary Compiler: Optimizes decision trees for speed and reliability, even under heavy workloads.

Transparency for Users and Engineers

We believe in making operations clear and actionable for everyone involved.

Engineering Monitoring: Provides real-time insights into performance and system health, allowing engineers to act quickly when needed.
User-Facing Progress Trackers: Track each order’s progress through the system, giving users visibility into key steps like validation, shipping, and tagging.

Looking Ahead: The Future of Automation

We’re not stopping here. Our vision includes:

AI-Powered Tree Management: Making it easy for non-technical users to create, manage, and optimize decision trees with minimal effort.
Modern Frontend Interfaces: Bringing cutting-edge design to deliver a seamless user experience.

By constantly innovating, we’re committed to empowering businesses with tools that scale with their ambitions.

This article was produced with the assistance of an AI language assistant to enhance its clarity, style, and structure. All technical content, ideas, and insights were provided by our team, ensuring the article reflects our expertise and vision.

The Challenge: How to assign different shipping carriers and services to 2k+ orders daily while lowering costs.

The Outcome: With the help of our advanced automation engine for assigning shipping services and carriers, the client managed to organize its logistics and supply chain, without the need for employees to remember every possible scenario. This resulted in a 0% chance for human error, while also lowering related costs by 20% and decreasing average handling time by 30%.

Details: Our client is selling its brands of textile products on Amazon (Fulfilled by a merchant), eBay, and Shopify worldwide, with the help of warehouses and shipping carriers in 11 countries.

He came to us with a problem – how to easily assign shipping carriers and shipping services to 1000+ orders a day without having the need to remember each rule for assigning carriers or the need to teach someone on their team to do that and remember every possible scenario.

The goal was to automate the selection process so that the chance for human error could be reduced to 0, while also decreasing cost and handling time while maintaining high-quality service.

We had a detailed discussion about the problem of our client and decided to implement a set of smart automated actions that helped alleviate the workload from the employees and made sure that the cheapest, yet fastest possible shipping service was consistently assigned to the orders.

What does the process look like?

Previously, the employees had sticky notes flying all over the place with written rules regarding how to filter the orders and assign the proper shipping carrier and service. This resulted in countless lost man-hours, hours that could be put to better use.

Now, Noverstock translates the business’ requirements into an automated rule, avoiding the need for employees to remember any of the specifics. The automated tool assigns the shipping services depending on the countries, weight, price, ”dangerous goods” flag, and marketplaces and could be changed at any point to include any other piece of data.

In addition to this, the customer can now easily change any part of the carrier assignment logic.

e.g. The business uses the cheapest untracked shipping services, but at some point, the deliveries are delayed or shipments get lost, which might deteriorate the health of the seller account. The business owner can contact Noverstock to implement a change to include tracked shipping service and thus make the best out of both situations. Whenever possible he makes sure to maximize its profits, and if the accounts are at risk, he quickly acts with the push of a button. All of this is immediately resolved and the Noverstock team makes sure the business needs are communicated and met on a daily basis.

This is just one example of how this rule could be used to better the operations and results of a business. It is one of our most used tools and it is unparalleled.

After implementing our automated action, the ease with which the system assigns the different shipping carriers and services completely eradicated the human error and the losses from inefficiently assigned shipping services, resulting in an immediate decrease in related costs by 20% and a decrease in average handling time by 30%.