Jenny Yule

Overview:

Amazon’s Outbound Communications team owned 17 automated push notification programs, with the intention of driving additional sales through product recommendations and reminders, powered by machine learning.

Our engineering team in India created the algorithms for the programs, as well as writing the customer-facing content.

The issue:
Due to the increasing consistency of negative customer feedback, culminating in an email from Jeff Bezos, there was a rush request to review and revise the algorithms, benchmarks, and content. The team paused these automated notifications, which represented a significant decrease in additional revenue.

My process:
As the only UX writer, I reprioritized my workload and joined with my partners in India in a rapid sprint process, working several early mornings and late evenings to expedite a solution.

I reviewed each program’s algorithmic strategy and its related copy to understand the disconnects, while also asking my engineering partners clarifying questions to ensure I understood their approach to each algorithm’s logic.

Next, I worked within the character limits of the push notification template to write copy that provided a clearer explanation of why the customer received the message. I also used a more neutral tone—even though engineering had tweaked some of the algorithms, I wanted to ensure the content could scale to whatever product might be shown.

Finally, I asynchronously shared the updates with senior leadership, using a table which showed the before and after, as well as the supporting algorithms. Once approved, I participated in final review and QA before implementation.

The entire review, revision, and relaunch of the programs took about a month, but my initial copy revisions were completed in five business days.

The result:
Because of the urgency to resume these campaigns, no copy testing was done. However, the relaunched programs, including revised content, had a 9% increase in click-through, a .08% decrease in opt-outs, and +55% increase in incremental sales.

Overview:
There was an impending launch for a product that would aggregate related content on one page, including customer ratings, third-party publisher content, and customer reviews. Its purpose was to help customers research and compare items before buying, without having to use an external search engine.

The issue:
The team had been using an internal product name, which had no contextual relevance to the final output and had legal copyright issues. They were struggling with a customer-facing name, so I was tapped to consult and provide them with some options.

My process:
I conducted a competitive analysis of similar experiences—including e-comm sites like Wirecutter and search engines. I also considered Amazon’s existing ecosystem, site architecture, and taxonomy.

I mined internal customer research to understand and avoid documented customer pain points, like confusion around existing Amazon product names.

The result:

My recommendation was do not “brand” the product with a customer-facing name. Instead, let its inherent value shine with ingress points to the aggregated content pages that are surfaced at relevant moments in a shopping journey. I presented my recommendation to the team and their leadership (view my presentation deck).

The team agreed that trying to brand the product with a customer-facing name wouldn’t do anything to create a positive experience—in fact, it might confuse customers even more, especially if there was no educational marketing campaign to support its launch.