FPT Software - Analyzing Employee Feedback to Improve Retention, Productivity Rates, and Workplace Satisfaction
Nhan DO Van Huynh Le Duy Tran Quang Minh Tan Le Hoang Khanh Nguyen Manh Nam
Organization: FPT Software
FPT Corporation is a leading global technology and IT services provider headquartered in Vietnam, with nearly US$1.3 billion in revenue and 30,000 employees in 26 countries. As a pioneer in digital transformation, FPT delivers world-class services in Smart Factory, Digital platforms, RPA, AI, IoT, Enterprise Mobility, Cloud, AR/VR, Business Applications, Application Services, BPO, and so on. The company has served over 700+ customers worldwide, a hundred of which are Fortune Global 500 companies in the industries of Aerospace & Aviation, Automotive, Banking and Finance, Logistics & Transportation, Utilities, and more.
Data Science for Good
Most Impactful Transformation Story
Most Extraordinary AI Maker(s)
FPT has more than 30,000 employees in 26 different countries, so understanding employees' satisfaction is one of the company's top priorities, as it directly impacts organizational success. Based on feedback from reviews or comments, we can determine which topics employees have issues with (e.g., salary, HR, leadership, management, working environment, facilities & infrastructure, overtime, training, and policies).
We collect ratings on each of these topics ranging from 0 to 5:
0: No mention
1: Extremely bad
5: Total satisfaction
This will help HR better understand what their employees want to change and whether those aspirations (reviews/comments) are appropriate so that the company and the business have a suitable solution to changing and retaining qualified employees, avoiding the loss of talent.
To prepare for this problem, we have collected data from various sources (personal websites, LinkedIn, etc.) with rating levels (positive, negative, and neutral) across 15 categories (salary, HR, facilities & infrastructure, etc.) to enrich the data over the years. During the preparation process, ETL manipulation (especially when correcting typos and normalizing data) is a difficult job because many of the staff here have a developer background, and there aren't many qualified members.
As previously stated, real-time data collection, maintenance, and reuse necessitate a clear, unified process that must be corrected as soon as any errors are discovered. We must maintain a team of 12 people to handle various tasks for this problem. As if by destiny, we discovered Dataiku, which proved to be our solution. It allowed us to find a way to leverage that existing ecosystem.
Our first impressions of Dataiku are that the interface is quite friendly, that data sets are simple to use and manage, and that data visualization is easy to do with dashboard creation, modeling, and feature engineering. What's more, ETL (extract - transform - loading) has become simpler with visual recipes on Dataiku, such as Pivot, Join, Prepare, Sync, and Split.
Our team of data scientists became more relaxed and focused solely on optimizing models (hyper-tuning parameters and run-time), developing recipes and plugins, and automating tasks. We saved a lot of time with Dataiku because it allows both data scientists and employees with a less strong background in data engineering to access, store, manage, and transform data before implementing Machine Learning algorithms.
As we have to collect data on a daily basis, managing data files can become extremely difficult if one of our team members is absent or resigns. Thanks to Dataiku, we only need three to four members to manage all of the work in this problem, from data collection and storage to checking workflow/schema consistency - ETL - modeling to the automation step and building the webapp.
Previously, we had to perform exploratory data analysis (EDA) with Python or Power BI, modeling with Python, and so on. We no longer need to make them separately, thanks to Dataiku — we can now use Dataiku's dashboard and use the code recipes or Lab feature from a visual recipe. This is extremely practical and user-friendly.
Business Area: Human Resources
Use Case Stage: In Production
Satisfied employees commit to working more and have higher retention and productivity rates. Workplace satisfaction can also be linked to other key factors in the context of OT (workload/stress), facilities infrastructure, supervision at work, and the balance between domestic activities and the work environment.
The first time the company launched this application was in April of 2022. So far, we have been able to see which employees are feeling overburdened with their work, whether the meals are qualified or not, and whether members of the same project are friendly with each other and satisfied with their PM or team leader. This prevents many employees from leaving for unknown reasons and also helps us save a lot of money when training with unclear and not particularly useful programs and courses.
It also helps us know which PMs and team leaders, as well as which HR representatives are abusing their power and oppressing their subordinates, so that we can have appropriate handling policies.
Value Brought by Dataiku:
This problem touches on multi-output classification, with training datasets of nearly 10,000 reviews a day. As a result, managing and integrating this data without Dataiku would be extremely inconvenient.
The number of FPT Software members who have obtained L2-Dataiku certificates is currently small (nine people), but they are spread across many departments and projects. We intend to grow to 40-50 members with similar skill levels in the near future.
The goal in the coming years will be to continue to upskill people with Dataiku to increase efficiency across more areas of FPT Software. As a result, using Dataiku to solve AI problems is a pressing need in this day and age. Compared to other platforms, our team prefers Dataiku because the documents and courses are very detailed, easy to understand, and not overly cumbersome. Plugin stores are also highly regarded and provide a wealth of useful information.
In my opinion, business analysts and data analysts may simply access data processing techniques like ETL and can even use Dataiku to operate machine learning algorithms. It can be stated that Dataiku will excite all users, whether low-tech or high-tech.