How Amazon is enhancing Customer Service with Write-Back

Amazon is a multinational technology and e-commerce company that was founded by Jeff Bezos in 1994. The company started as an online bookstore and has now grown to become one of the largest and most diverse online retailers in the world.

Amazon has expanded its reach beyond e-commerce and has ventured into cloud computing with Amazon Web Services (AWS), streaming services with Amazon Prime Video, and the development of proprietary devices such as the Kindle e-reader and Echo smart speakers.

Delving deeper into this story, let us focus on a special division within Amazon’s great universe.

Amazon’s Devices and Services division has the responsibility of creating and developing products that offer more value when combined than they would if they were separate. The Division’s goal is to achieve ambient intelligence, which involves the use of artificial intelligence (AI) to integrate devices and services in a way that generates more value than any single product could on its own, whether at home or on the go.

Among many services from Amazon’s Devices and Services division, we find Alexa. The Alexa service continues to improve as new features and experiences are added, making it more knowledgeable, proactive, and natural to interact with.

It is not surprising that Alexa is one of the most widely used virtual assistants globally and boasts millions of users worldwide. Consequently, Amazon maintains a large-scale customer service team to support these users.

Within the Amazon D2 Customer Service division, the customer service analytics team plays a pivotal role. Their responsibility encompasses all analytics processes in this domain, with the ultimate goal of providing insights that enhance customer service. However, conducting analytics at such a massive scale presents challenges. Specifically, offering feedback to address exceptions or performing qualitative analyses becomes complex.

Previously, when leveraging analytics with Tableau, the customer service team had to switch contexts, relying on separate tools like Excel and SharePoint to implement these processes. This cumbersome transition consumed valuable time and hindered the proper standardization of inputs.

Discovery & Setup
To address these pain points, Amazon sought a solution that would allow for interactive dashboards and streamline the entire analytics process within a single platform. In this success case, we will take a closer look at how this product helped Amazon overcome these challenges and the benefits it provided, and how Amazon is enhancing Customer Service with Write-Back.

The discovery of Write-Back by the analytics team through an online search was a pivotal moment. As they delved into the documentation, it became evident that this product aligned perfectly with their requirements. However, like any new endeavor, there was a learning curve. Yet, the setup process unfolded seamlessly, and before long, Write-Back was operational, empowering the team to streamline their analytics workflow.

Use case: Customer Agent Performance
Effective customer service in a call center relies heavily on the performance of customer agents. Team managers play a crucial role in evaluating this performance and making decisions about each agent’s development cycle. These development cycles provide targeted guidance and improvements. While automation algorithms can handle most aspects of cycle management for each customer agent, there are situations where human intervention is necessary. This critical data about people cannot be solely entrusted to automation, especially when dealing with exceptions related to external factors. For example, consider a scenario where a customer agent worked only part of the month, resulting in lower performance metrics. Such fluctuations are normal, but they require manual adjustments.

Team managers step in to correct these exceptions. However, any changes made must be approved by human resources and leadership to ensure consistency and fairness. Privacy compliance is also a priority. To address these complexities, specific dashboards have been created for each role. These dashboards automatically filter data, creating a hierarchy of exceptions and approvals.

In summary, finding the delicate equilibrium between automated processes and human judgment is crucial for maintaining a high-quality call center experience while upholding privacy and fairness. Amazon successfully constructed this intricate process using a Tableau dashboard package, seamlessly integrating Write-Back. Not only did Write-Back standardize exception inputs and directly feed data into analytics dashboards, but it also facilitated the implementation of an approval mechanism. This entire process now resides fully within the analytics platform, thanks to the flexibility afforded by Write-Back.

Use case: Contact Mining
Write-Back was available at Amazon and deployed successfully for the first use case, so when the analytics team had to improve the contact mining platform, it was easy to get approval from stakeholders to use Write-Back in this use case as well. Contact mining is all about having quality analysts going through contacts, which can be email messages or conversations, and providing comments or suggestions that can be used for coaching customer agents. In a support case, there are situations where the customer might be unhappy, and this has nothing to do with the customer agent’s performance hence the importance of determining if the contact is operations controllable or not controllable to provide accurate advice. The mining process is implemented across many countries and having Excel register these decisions prevented proper standardization requiring a lot of manual work. By integrating Write-Back on the Tableau dashboard Amazon was able to create a centralized repository for manual inputs. This not only made the submission process much more streamlined but also opened the possibility of establishing benchmark KPIs (key performance indicators) within a certain country bringing a whole new set of analyses in terms of customer agent behavior, for instance, what is the percentage of friendly conversations you expect to have within 6 months on a particular country. This enabled the establishment of more accurate coaching and further improved customer service.

Expanding further to other use cases
The Amazon D2 team keeps expanding Write-Back usage into other domains, following similar patterns to those achieved on contact mining. Stakeholders have seen how it works and want to leverage it on mode use cases where users can provide comments about KPIs highlighting exceptions and contextualizing it and with this achieving a collective qualitative analysis directly stored on the analytics platform and ready to be analyzed. The latest examples are use cases regarding Operation Intelligence and compliance but much more will likely arise.

Customer Agent Performance

Write-Back brings the automation of the performance management process for the first time in D2, which saves the analyst’s bandwidth by 75%. Also, thanks to this initiative and the capabilities brought by Write-Back, the prediction of cycles for agents is now automated.

This overall project improved D2AS worldwide key metric by 5%, saved operating costs by $5.8M, and contributed to a contact reduction of 70K.

With over 5,000 unique users and 200,000 views across all performance management dashboards, we can confirm the significant role of this joint initiative between Amazon and Write-Back in improving business processes.

Contact Mining
The primary aim of this initiative was to automate the process of contact mining by using Write-Back for contacts that were not present in D2. This objective was achieved with enormous success by improving the previous process. Earlier, different organizational units (OU) used different processes to mine contacts using Excel/SharePoint.
The mining of data also helps Amazon D2 to understand the non-controllable situations faced by their associates, and some of these situations can be escalated to business and product teams as opportunities.
In addition to the original objective, this initiative not only automates the process but also provides a standardized data source that is widely used by several Amazon OUs for various use cases.

As a result, using Write-Back to solve this Amazon D2 requirement led to a 10% (4-hour) reduction in bandwidth for each L5 quality analyst per week.

Overall, Amazon is enhancing Customer Service with Write-Back, and is optimizing costs and improving key metrics.

“Write-Back transformed our approach to customer agent performance, introducing automation and saving our analysts’ bandwidth by 75%. The results speak for themselves – a $5.8M cost saving, a 70K contact reduction, and a 5% improvement in key metrics. A remarkable achievement!”

Radhika Malviya, Senior Business Intelligence engineer at Amazon

See other Success Cases