CASE STUDY High Tech

MACHINE LEARNING SOLUTION HELPS PRINTING COMPANY SAVE MILLIONS IN SHIPPING COSTS

Over the last decade, Saggezza (an Infostretch company) has had an ongoing relationship with a large printing and distribution company. The customer produces and mails catalogs, retail inserts, magazines and books for its clients, an undertaking that sees more than six billion physical items shipped every year.

Distribution and logistics have long been a focus for digital transformation, even more so when you consider the volume of physical mail, marketing collateral and commercial material that is sent out every day. And while bulk mail is the result, distributing printed B2C assets is often impacted by how much they cost to ship.

As part of its commitment to lower overall postage costs, the company encourages its customers to join a co-operative mail program. This program is a key part of the process, allowing the company to combine participating customer orders in one batch and distribute them more efficiently across the delivery ecosystem.

However, the sheer volume of physical products often meant that shipping costs exceeded printing costs. In addition, the batches of printed material had a staggering number of postal configurations, all of which increased both the number of hours spent in assembling the batches themselves and the varying shipping costs associated with timely distribution.

  • An iconic US retailer

    Prints and distributes 6 billion catalogs and magazines annually

  • multiple stores in the U.S

    ML saved over $70 million in operational costs

  • $11B in annual revenue

    Delivered solution won industry award for AI excellence

test automation in retail

Faced with this challenge, the company tasked Saggezza with developing a solution that could not only reduce the total shipping costs but also optimize the batched materials into workable configurations for distribution. Saggezza’s solution to the problem was to introduce data and analytics into the mix, with an internal team leveraging machine learning to solve the demonstrated pain points.

Thanks to the 10-year involvement with this company, the delivered solution has already been beneficial. By using artificial intelligence and machine learning to solve the configuration and batching problems, the client will be able to save more than $70 million, while an optimized configuration can be identified in less than 90 minutes.

In addition, the innovative use of AI/ML in a real-world problem saw Saggezza receive an Artificial Intelligence Excellence Award from the Business Intelligence Group.

the Results

Key Outcomes

Demonstrated Cost Savings
Demonstrated
Cost Savings

AI/ML integration will save the company more than $70 million in operational costs

Optimized configuration process
Optimized
configuration process

Optimal material batch size and configuration for distribution identified in 90 minutes or less

Co-operative mail program
Co-operative
mail program

Maximized savings through ML will allow company to offer competitive fees and increase customer satisfaction

Co-operative mail program
Efficient process = new customers
+ increased growth opportunities

Automation alleviates distribution delays and batch mis-configuration

Our methodology

how
we did it

Infostretch works with companies across the digital lifecycle.

Go Digital
Go Digital

Accelerating the delivery of new digital initiatives with confidence

Be digital
Be digital

Creating the infrastructure and foundation to scale digital initiatives

Evolve Digital
Evolve Digital

Leveraging data and analytics to continuously improve digital delivery processes

The challenge

REDUCE SHIPPING COSTS, INCREASE OPERATIONAL EFFICIENCY IN BATCH CONFIGURATION

Commercial mail delivery is often offered to companies at reduced rates by third parties such as the U.S. Postal Service or specialized vendors. However, those costs can quickly mount up when the number of items being batched and dispatched run into the thousands per consignment.

In the case of the printing and distribution company, the shipping costs not only formed a large part of the operational expenses, but also exceeded the total printing costs of a project. As part of its cooperative mailing program, the company worked with a third-party vendor to deliver the printed materials to customers, a partnership that relied on the discounts that said vendor could provide for presorted bulk mail.

The Problem

These printed materials ended up being a variety of sizes and batch configurations. This meant that the company had to continually set its onsite packaging equipment to create the bundles required for the vendor. In addition, the number of potential configurations made it almost impossible to come up with the optimal batch in not only a timely fashion but also one that could take advantage of the discounted rates.

Saggezza’s brief was therefore simple; find a way to shrink the number of possible configurations down to a smaller, workable set of optimal ones and select the best configuration from this smaller set, considering both equipment and timing constraints.

The Solution

INTEGRATION OF MACHINE LEARNING INTO CONFIGURATION AND DISTRIBUTION

The data and analytics team determined that a custom machine learning solution was the best way to achieve desired outcomes.

Data being generated by the batching process would follow certain patterns. The key element was to find those patterns and apply them to the optimization metrics required. In this case, the team used unsupervised learning algorithms to find and group unusual patterns – for example, the constant equipment configurations and the batched material cost requirements.

Unsupervised learning

Unsupervised learning is the ideal way to integrate an automated process into logistics as it relies on methods that do not have “labeled’ data. In other words, it looks for patterns that do not rely on human input. The algorithms can then group items such as equipment configuration that generate cost savings for distribution and apply that to the items to be batched.

This approach allowed the client to configure the packaging equipment to group batches of printed material for third party distribution. The algorithms work in tandem, feeding each other with outputs that optimize the solution and look for ways to improve it beyond the original brief.

For example, there were two types of algorithms working together – one genetic (simulating natural selection) and one hierarchical clustering algorithm with structural constraints (introduced to group the printed material into the most optimal for mail carrier routes). This tag teaming of algorithms thereby allowed the company to optimize equipment configuration and cost savings in one single process.

Partnership Platforms/Software
Snowflake

Azure

Tableau

Python

PROBLEM + CHALLENGE = SOLUTION

As distribution and logistics becomes ever more digitalized, companies that can optimize machine learning capabilities to reduce the time spent batching physical products will be a vital part of the ecosystem.

The solution for this printing and distribution company was designed to maximize the opportunities that automated processes can offer, with the optimized configuration time reduced to less than 90 minutes.

That leads to two defined results:
time spent on equipment configuration and material batching

The time spent on equipment configuration and material batching is more attuned to the requirements of the customer.

operational cost savings generated by machine learning and automated processes

The operational cost savings generated by machine learning and/or automated processes means that the company can offer its customers competitive pricing for those that join its cooperative mail program and, importantly, increase customer satisfaction via a more efficient distribution strategy.