18 months. That’s how long it took the Statistical Office of Republic of Serbia to publish its official census results before. This time, it delivered preliminary census results in just one month, and official results in six months. This was made possible thanks to the team adopting Microsoft solutions such as Azure, SQL Server, and Power BI. Not only did it help the Office conduct a paperless and efficient census process; it also ensured top data security, enabled real-time monitoring, and improved data quality.
The Statistical Office of Republic of Serbia is responsible for conducting population census and collecting important data on geographic, demographic, socio-economic, and ethnic-cultural characteristics. With a central office in Belgrade and 15 regional data centers, the Office plays a vital role in Serbia's data management and sharing. Its presence across Serbia enables easier collaboration with other governmental bodies through facilitated data flows.
One of the essential tasks undertaken by the Office is the national census of the population, which is conducted every 10 years. The census aims to provide complete, high-quality, and internationally comparable statistical data, aligning with EUROSTAT guidelines and global statistical benchmarks, that are vital for planning social and economic development. Miladin Kovačević, the CEO of The Statistical Office of Republic of Serbia, proudly explains, “This was the first digital census in Serbia, carried out entirely paperless using Microsoft tools.”
Before this, the Statistical Office faced a significant challenge in conducting the census: its legacy on-premises equipment couldn't handle the required number of concurrent users, and there were security concerns. The Office decided to use a hybrid cloud solution, as Kovačević explains. "We use the cloud only for non-identifiable, aggregated data and encrypted local storage for personally identifiable information. This approach enabled us to streamline maintenance while ensuring data privacy and security." This enabled the Office to focus on its core tasks of collecting, aggregating, and analyzing data, ensuring a successful digital census.
A comprehensive digital solution for Serbia's census
The Statistical Office’s new digital solutions transformed its census process, starting with its in-house software, IST. The program was created exclusively with Microsoft tools such as SQL Server and .NET technology, enabling seamless data collection that results in higher quality data. Kovačević emphasizes, “The SQL Server is the heart of our system. It allows us to encrypt data and protect it against unauthorized access.”
Using Microsoft Azure VMware Solution, Azure Arc, and Azure Virtual Desktop improved accessibility and data management. For data storage on laptops, the Office leveraged Microsoft LocalDB to secure sensitive data that’s stored on local devices. Additional solutions enhanced Office’s security posture and compliance with standards, including Microsoft Sentinel, Azure Monitor for on-premises security, Web Application Firewall to secure access, and Microsoft Defender for Cloud, for SQL, and for Servers security.
To compile data for the census, the Office sends enumerators to survey residents across the country. For the first time, these employees could obtain and share real-time data analysis, via Power BI. This also enabled Kovačević and his team to monitor the performance of enumerators and allocate resources effectively. “In this census, there were 15,000 enumerators and 2.5 million locations, compared to 40,000 enumerators across 2 million locations in the previous one,” explains Kovačević. “Because we can monitor the progress of every enumerator through Power BI, we could optimize workloads instead of simply allocating more people to the census.” Live maps embedded in Power BI offered visual insights. The Office also created a Power Apps-based ticketing system which enabled the regional centers to handle IT issues raised by enumerators working in the field, further enhancing fieldwork management.
Faster data encoding and more accurate results
Having made data collection and storage more effective, the Office then needed to speed up data encoding and analysis. Since Kovačević and his team create machine learning (ML) algorithms on a daily basis, they built new ones to code open-ended responses in the census questionnaire.
Answers to items like occupation and nationality need to be normalized, and the ML algorithms helped automate and reduce data processing time. “Before, the process of manually encoding answers to open-ended questions from seven million paper responses was done nine months after the forms were filled. We had to manually enter data with an Optical Character Recognition (OCR) scanner,” recalls Kovačević. “Now, it is done paperless and in two hours, thanks to machine learning technology and cloud solutions.”
With all the new tools, the Office achieved impressive results, such as obtaining preliminary census results in just one month. Official results were published in six months, compared to the almost two years it had taken in the previous census. The error rate was also reduced to a mere 1%.
“Before, the process of manually encoding responses to open-ended questions from seven million paper responses was done nine months after the forms were filled. Now, encoding is done paperless and in two hours, thanks to machine learning technology and cloud solutions.”
Miladin Kovačević, CEO, The Statistical Office of Republic of Serbia
The success of this first digital census in the Republic of Serbia is evident. “The feedback from everyone was excellent!” Kovačević exclaims. “Everything was done smoothly and on time, which was the most important thing for us: to have proper data quality at the right time. The method we used for the last census now feels like something from the stone age. It’s so different that it's almost incomparable.”
The new solutions have not only revolutionized the census process but also positively impacted the entire office, making the technologies reusable for future surveys. “Just one year after finishing fieldwork for the census, we have another: the agricultural census. It’s very complicated,” explains Kovačević. “We plan to redeploy the new solutions for it. We are already prepared, and we expect things to go much smoother and much easier thanks to all this experience.”
Follow Microsoft