Connections provides Amazon employees a confidential way to provide feedback on their work experience to help shape the future of the company.
In 2014, Jeff Bezos visited a Fulfillment Centre in Kent WA and saw that associates answered a few quick questions about their work day as they exited the security turnstiles. He asked if there was a way to automate and scale this concept.
The MVP launched in US-based Fulfillment Centres in 2015 with questions to understand Safety, Leadership Behaviour, Job and Manager Satisfaction from hourly associates.
The launch was successful with 98% engagement and improved safety standards. It proved the concept and a small experimental team was tasked to scale the experience for all Amazon employees.
In 2017, Connections launched to ~400k corporate employees. The daily question app appears after employees log in with their Amazon account credentials.
We analyze big data with AI, ML and NLP to learn, understand and predict behaviour in an effort to impact positive change at the company, around the world.
Connections ML uses a genetic algorithm to encode possible solutions into a population of chromosomes that evolve through mechanisms like selection, mutation, and crossover. This generates high-quality solutions for optimization problems in the product in the form of actionable insights and recommendations.
Today, >1MM employees across every major organization pillar start their day with a daily Connections question.
Privacy and security are key. Only managers with teams that collect 4+ responses to a question can see their results in the Reporting tool.
Less than 40 researchers worldwide have access to individual responses. Connections data is otherwise tokenized, aggregated, and accessed on a team-level basis.
Managers use the Reporting tool to analyze question results and receive actionable insights and recommendations based on trends over time.
I redesigned the Reporting tool in 2020 to accommodate emergent themes related to Covid-19, obstacles to productivity WFH, and virtual new-hire onboarding. This timely data was critical for Amazon’s SVP of HR to change policy.
Sub-themes are drivers of overall theme trends that provide additional context for managers to deep-dive cuts of data by population and question type.
Root cause questions are unscored and allow employees to write open text feedback. This mechanism provides clarity for managers to take action with.
Teams are encouraged to review and discuss Connections results weekly and assign owners to improve themes they care most about.
Connections receives >300MM annual responses from employees. This massive data lake makes it possible for ML models to accurately predict events based on behavioural and attitudinal changes in employees over time.
Today, Amazon VPs and SVPs rely on Connections Reporting to understand and make science-backed organizational changes at-scale.