21 Dec 2019, 10:00 — 10 min read
Background: Every modern retail business dreams of becoming the ‘perfect’ business. A perfect business is one that is profitable, scalable and one with error-free operations. Is it really possible? Can there be a business that has zero errors? Amit Verma, Director-Engineering, BigBasket.com, one of India’s biggest players in the online supermarket space believes that zero-error businesses exist and that BigBasket is one of them. Here’s why…
Supply Chain Management and Delivery Operations are all about decision making. There are some macro decisions as well as countless micro- decisions that need to be made on a daily basis. For example, when an order is placed, the decisions to be made include ‘when should picking start?’, ‘which is the best route to take?’, ‘which picker should pick this item?’, ‘should all the items of order be picked together?’, ‘can chilled items be picked last?’, ‘should this order go by bike or van?’. This decision making happens both in the virtual world with custom software as well as the physical world by people running the process for BigBasket on ground.
BigBasket.com currently operates in 25 cities and has more than 70 warehouses. The scale, requires proper utilization of data. Only portions of data which make sense to the concerned person need to be sent. The nature of operations, especially for BigBasket, are highly dependent on local conditions and so items can’t be standardised. For the data to support decision making reliably, BigBasket uses tailor-made software, which is flexible (configuration- driven), has an inbuilt Retry and Recovery mechanism, allows forecasting capabilities, provides easy access to reports, offers custom dashboards and monitors and sends alerts.
According to research by Deloitte, 79% of companies with high performing supply chains achieve revenue growth greater than the average within their industries.
“There is a lot of movement of data across our software systems. Software being software does fail. That is why we have built self-healing systems for certain scenarios, where data, if not correct, gets auto corrected in next exchange. On the process side too, if a person has to take a decision, say a picker wants to know what item he is supposed to pick, when he should pick it, should he pick items for orders together. Everything is provided right there at his/her fingertips,” explained Amit.
Driven by technology the error- free operations has the following aspects: Customer, Inventory Data Management, Efficient Picking & Sorting, Improvement through Reports & Feedback, Monitoring & Alerts and Manual intervention.
Also read: XXL Growth
“BigBasket is a customer centric company and all its processes, solutions, results are geared towards making the customer happy, first. Error in operations starts from what you promise to customer. So we don’t take orders if an item is not available or when we don’t have delivery resources to deliver the order at a particular time-slot,” said Amit.
Also, it’s important to understand what zero-error operations really means. “Error in operation can mean many things to different people but from a customer pointof- view, it means two things—all items are delivered and that there is no delay in delivery. As long as the customer is getting all the items delivered on time, s/he is happy,” explained the Director.
For most of its locations, the delivered items are close to 100% and even on time delivery is ~99%. This, despite the fact that Bigbasket does not use drones, conveyor belts, pods or robots. The minimal error in operations stem from the following core principles:
Also read: Customer is the king and the kingmaker
Items a customer reserves on Big- Basket.com become an order if the payment succeeds. If it fails, they get unreserved and the order is abandoned. The resources for delivery such as van, bike, executives are estimated weeks in advance and the capacity number is added to the system. If the entire capacity for a slot is exhausted, it becomes unavailable for booking. Here, proper inventory management becomes critical. Incorrect inventory data is one of the main causes of items not being delivered to customers. From a software point-of-view, inventory management is about addition and subtraction of inventory data. The physical inventory sits in the warehouse but inventory data exists in multiple systems due to their distributed or micro-services architecture.
The challenges that any large scale distributed warehouse operations face is having consistent view of data across various systems. The common approach is to batch the inventory related operations locally and then push the batch to other systems, particularly to the Order Management System. Until a few months ago, BigBasket too followed the same procedure. While the frequency of this sync up with other systems could be increased, there would always be a time gap during which the data would not be consistent. This led to more orders getting placed for less inventory. “We built a Custom Inventory Data Transfer Solution, which enables all inventory changes to propagate at the same time to other systems. In case the change is missed in one call, it gets corrected in the next. Additionally, while the data gets self-corrected, the rouge data is stored and sent for further debugging. This has improved our fill rate significantly,” informed Amit.
When the inventory of, let’s say, apples on the shelf is same as what the customer sees on the app, the chances of error reduces. There are always two kinds of data here: Warehouse Management Systems inventory data and data that customer sees as available. Ideally, these two should be in sync. However, there are times when they are not.
To address this, BigBasket built an algorithm which achieves almost real-time synchronization, while also capturing discrepancies. It also ensured that data movement did not block local operations. This optimization significantly improved the efficiency of pickers.
The system enables better utilization of data for decision making in operations. It also provides a dashboard for Delivery Executives bringing both functionality and ease into the delivery process. There is a historical Reporting System with a feedback loop which is used by engineers to further improve the algorithm by fixing the discrepancies in the process.
The earlier process assigned the picker a set of orders, which s/he then picked sequentially. So if each of the five separate orders had apples in them, the picker would go to the apple location five different times. In the new system, all orders of a picker having apples are grouped together. Now, the picker carries a cart and picks apples for all the orders in one go. This means that the picker has to go to one item location only once, saving time and effort. It has also led to lower attrition and higher earning. And has saved 30 minutes in the whole supply chain process.
Another improvement is in the grouping of orders. The older process had the picker pick orders and place it in a common area. A segregator then grouped the orders based on routes. This was a manual process that would take 30 minutes for a given slot of around 60-70 orders. Also, there would be occasional mistakes in segregation. With the new system, this process has been eliminated, which not only reduced manual errors but also the cycle time by 30 minutes.
Also read: Retail out of the box
For a company that has around the clock operations, monitoring and alerting are pivotal. “Some critical functions such as placing stock orders or computing sales forecast for the next day happens at night. Hence, it is extremely critical to ensure that if there is any break of pipeline at night, the right folks should be alerted,” explained Amit. The company has built a fully automated software that shoots off Business Metric Alerts to let the business know about failures if any so that they can be mitigated in time.
However, despite the best of code being written by the best of engineers, there are bad days and nights, when the entire tech team somehow misses the alarm. At such times, a delay of even 10 minutes could result in a non-zero error operation. Even when the tech team debugs the issue, the operations can’t wait. That’s why, we keep manuals to Key Business or Operations to follow during such times to ensure business runs as usual,” said the BigBasketeer.
With technology, processes and plain common sense BigBasket seems to have achieved the impossible goal of zero-error operations.
Article by Amit Verma published in STOrai Magazine.
Disclaimer: The views and opinions expressed in this article are those of the author and do not necessarily reflect the views, official policy or position of GlobalLinker.
17 Jan 2023, 14:23
7 Dec 2022, 14:48
Most read this week