Media

Perfect Corp(玩美移動) App

Customer Background

Perfect Corp (玩美移動) is the best all-in-one photo editing app powered by AI. Instantly elevate your snaps with AI object removal, background removal, aesthetic photo filters, collages and frames.

Challenges

• • company struggled to keep up with frequent product updates, often lagging behind in processing new and modified product data.
  • With billions of clicks per day (equating to 200GB), each query took a considerable amount of time to process, leading to significant data storage and management costs.
  • The existing UI lacked flexibility and couldn’t display individual user profiles and behaviors. Any new data dimensions required IT intervention, but the development was tedious. Moreover, the streamed data hadn’t undergone cleaning and validation, casting doubt on its accuracy.
  • These issues resulted in operational decision-making glitches and a lack of confidence in the accuracy of their data.

Customer Requirements

• • Establish a real-time streaming video data platform.
  • Develop predictive models to forecast traffic volume for the next 5 and 15 minutes.
  • Present the predictive models and analytical reports to assist in strategic decision-making.

Solution

• • On BONA, business/marketing personnel who are unfamiliar with SQL can quickly view result charts using a drag-and-select method. The platform can promptly receive and convert key-value semi-structured data into structured data as product dimensions are added or modified due to updates.
  • BONA can process up to 3 million records per second, displaying them in real-time on its dynamic dashboard.
  • The system supports cluster and time-series query methods, delivering results within tens of seconds for each query.
  • Users can directly click on statistical charts to drill down into individual user profile details, such as gender, residence, age, etc.
  • On BONA, users can freely adjust various dimensions and conditions, swiftly observing the results after different permutations and combinations of dimensions. The platform enables rapid data dimension inspection and validation, automatically cleaning any anomalous data.
  • The system compresses data to one-seventh of its original volume, significantly saving on data storage space and costs.

 

Results

Dynamic real-time data – emphasizing speed of transmission and data scale (Velocity and Volume)

• • Data is continuously imported into the BigObject database at a rate of 300,000 events per second.
  • More than 20 billion events are imported daily.
  • Monthly data volume exceeds 5TB.

Semi-structured Data – Data Types (Variety)

• • Semi-structured data.
  • Diverse data sources and multi-dimensional data.
  • Using BigObject’s distributed (Cluster) architecture.