Improved Order Fulfilment Efficiency & £6M+ Annual Savings

Introduction

During my career, I had the privilege to work with some of the brightest minds on many innovative projects. Out of all, if I had to pick one of my ideas that got implemented and had a huge impact, I would choose the “Picking Application” project at Tesco. Before we dive deep into the problem statement and the solution, I will first cover essential domain knowledge and provide high-level context.

Domain Context

Tesco is the 3rd largest retailer in the world by revenue and fulfils around one million online orders every week during peak demand periods like Christmas and other holiday seasons. I was part of the Customer Fulfilment team. Once customers place orders through Tesco’s online website or mobile application, it is my team’s responsibility to process these orders and deliver them to the customer.

As part of the order-fulfilment process, my team creates thousands of Pick-Trips (list of order items to be picked) every day and sends them across 3,000+ stores and warehouses. Every week, 100,000+ personal shoppers (Tesco employees who pick your order items) use “picking devices” to request Pick-Trip details. These are hand-held devices similar to mobile devices with an inbuilt scanner that can be used to scan product barcodes.

Pick-Trip Lifecycle

• 1.Personal shopper scans the “Start-Pick-Trip Barcode” placed in the loading area. The picking device receives Pick-Trip details.
• 2.The personal shopper then takes a trolley and enters the shopping floor to pick the order items in sequential order as displayed on the device screen. On average, a single Pick-Trip contains around 40–60 items.
• 3.Once the personal shopper picks all the order items in the Pick Trip, they exit the shopping floor and return to the loading area.
• 4.They then scan the “End-Pick-Trip Barcode” placed in the loading area and it marks the end of the pick-trip. At this point, the device displays some stats about the picking trip.
• 5.One of the key metrics that is important to Tesco and the personal shopper is the Picking Rate—i.e. number of items picked per hour.
• 6.The personal shopper then scans the “Start-Pick-Trip Barcode” again to collect a new pick-trip and repeats this picking process until the end of the shift.

Project Context

Tesco’s official hardware vendor/supplier for the picking devices rolled out a new version of picking devices with advanced features like Android OS and an improved scanner with a faster scanning rate. As part of this new device rollout, the vendor stopped providing support for the old picking devices used by Tesco across its stores. We had 18 months to migrate to the new picking devices and the business case to migrate to new devices received a go-ahead.

We built an Android application from scratch with all the must-have functionalities. We rolled out the initial prototype (beta version) to 10 stores to gather feedback and monitor for errors. The beta version was working as expected and we received a request for a new experimental feature to display a leaderboard—rankings for all the pickers in the store based on Picking Rates (number of items picked per hour).

The store managers requested this feature because it enables them to reward personal shoppers (pickers) based on their performance (Picking Rates). It also provides an incentive to personal shoppers to work harder and eliminates slack. The feature request was approved, and the Leaderboard feature was developed and rolled out across 10 stores.

After monitoring for 4 weeks, we noticed that the fastest pick-rate improved across all stores and store managers were happy with this new feature. However, the average pick-rate declined by a small percentage, and it led to unhealthy competition between personal shoppers and a toxic workplace culture.

Right around this time, I transitioned from the lead developer to Technical Program Manager role and was responsible for rolling out the application from 10 stores to 3,000+ stores. I was skeptical about the experimental leaderboard feature and opposed its roll-out to the next 100 stores. However, based on positive reviews from the store managers, the decision was made to rollout the feature to the next 100 stores.

Problem Statement

After the rollout to 100 stores, we noticed a problem. The average pick-rate went down across 80% of stores and the feedback received from the personal shoppers—who are the real customers for the picking application—was negative. There were complaints about unhealthy competition between the personal shoppers.

While the top 10% of pickers did well, the remainder of the personal shoppers who were not at the top of the leaderboard got de-motivated and started performing badly. We were now faced with a choice of whether to roll back the experimental feature (leaderboard) or go ahead with it. The store managers were excited about the feature, but the personal shoppers (the real end users of the application) were not.

Solution

I had conducted a series of brainstorming sessions with the team and business counterparts and gathered feedback from personal shoppers. After analysing all the data points, I proposed an innovative idea that would provide the same benefits as the leaderboard feature minus the unintended side effects.

Instead of displaying a leaderboard based on picking rates (number of items picked per hour), we could display a real-time “BRAG Bar” during the pick-trip process.

What is the BRAG Bar?

We receive real-time updates on items getting picked by the personal shopper during the pick-trip. We then use this data to calculate their pick-rate and classify it into one of the following categories:

• Red — If the picking rate is less than 20% of the average picking rate for the store. (PR < -20%)
• Amber — If the picking rate is between -10% to -20% of the average picking rate. (-20% < PR < -10%)
• Green — If the picking rate is between -10% to +10% of the average picking rate. This is the ideal picking rate. (-10% < PR < +10%)
• Blue — If the picking rate is greater than 10% of the average picking rate for the store. (PR > +10%)

At the top of the picking application screen (fixed header component), one of the below components glows based on the personal shopper’s current Picking Rate (PR) compared with the average picking rate for the store.

BRAG Bar — Picking Rate (PR) thresholds

Result

This new feature (BRAG Bar) was rolled out to 10 stores and monitored for 4 weeks. At the end of 4 weeks, we noticed that the average picking rate increased by 12% across all stores.

As the personal shoppers could view their picking rate BRAG live during the pick-trip, the personal shoppers who were in the red zone cut down slack and improved their performance to fall in the Amber or Green zone. 90% of all personal shoppers who were previously in the Amber zone improved their picking rate to fall in the Green zone. Personal shoppers who were in the Green or Blue zone performed consistently to be eligible for rewards and bonuses.

This also eliminated the problem of unhealthy competition and toxic workplace culture among personal shoppers, as the leaderboard feature was removed. The store managers were happy too, as they could still track their best performers. The feature was eventually rolled out to all 3,000+ Tesco stores along with the new picking application.

As a result of my innovative solution and its implementation, the average pick-rate went up by 12%. It led to an average annual savings of 2,000 pounds per store, which translates to a combined annual savings of 6 million pounds for 3,000+ Tesco stores.