A PoC Project: Customer Conversion Data Exploitation With Camera At Retail Stores
A PoC project that proves the ability to collect data of customer going in/out-store and conversion rates using existing cameras at the retail chain.
A leading retail company in Vietnam with more than 500 stores, is looking for more opportunities in tracking the amount of customers and conversion rate of customers visiting stores turning into customers buying products/services, while at the same time, relying on that data, to devise a strategy for allocating staff resources accordingly.
Data is a valuable source of information that if businesses know how to collect and utilize, it can bring tremendous value to the organization. By analyzing the data, the retail corporation can monitor and better understand customer behavior, as well as offer solutions that are appropriate for the situation.
Seize the opportunity: Is it possible to monitor the number of customers at stores?
Enterprise wants to capture and build new technology solutions based on existing values. Through working process with the business, we realized that we could further improve the visitors counting solution by using cameras and computers available in stores. In-store cameras are of uneven quality and have only been used for security surveillance. We have studied the application of human counting technology based on artificial intelligence basis from existing cameras. This technology makes counting visitors in and out of the stores to become more accurate.
To prove the feasibility of the technology initiative, we work with client to carry out the Digital PoC (Proof of Concept) project, which helps evaluate and assess return on investment (ROI) and the feasibility level quickly in the early stage, while at the same time, save resources for business.
Identify the core strategic requirements and goals
It is essential to identify the core strategic requirements and goals to be able to build appropriate products/services. By doing market research, applying experiences of technology experts as well as understanding customers demands in utilization of existing cameras, FPT Digital plans core strategic requirements and goals in the development of customer counting system to ensure:
- Cost-saving in building customer data for business
- Optimal utilization of currently available resources (computers and cameras) in stores
- Ability to connect and retrieve data directly from surveillance cameras or collect information from stores’ portal
- Ability to configure, support, integrate, upgrade and expand easily
- Counting tool’s speed optimization so that computers at stores can run the counting tool
- Algorithm optimization to improve people identifying model to reach the accuracy level of over 80% in the number of visitors in/out of each store by hours, distinguish employees and customers, and be applicable in stores with different layouts
- The ability to report automatically store locations with the numbers of customers at different times
Implementing the project of counting visitors in and out of stores with FPT Digital
FPT Digital works with businesses and supports implementation:
- Alter algorithms and instruct installation at stores
- Implement for the first time, train at each store and make adjustments
- Test on old videos (provided by the enterprise)
- Optimize algorithms and models to achieve the best performance
- Package and integrate with the system of the business
Testing validates the viability of the solution
Within 5 months, we conducted tests on 500 stores of the business. The result is that the accuracy rate of counting customers in/out of stores stays above 92%. Successfully proved the feasibility of the idea, this tool continues to be refined and optimized across all stores of retail business in Vietnam.