Behind-the-scene of the 2022 Digital Lookback" Campaign on BAEMIN

Tech news
Tech release

2022 marked a vibrant year for BAEMIN's engagement campaigns, showcasing diverse gameplay and creative concepts. In particular, the 2022 Digital Lookback campaign performed particularly well with an impressive in-app CTR of 3 times the average rate*. Let's delve into the technical elements that contributed to the success of this campaign. 


From idea


Harnessing the power of data, our Data team continuously shares ideas and proposes enhancements to develop BAEMIN's products and elevate the user experience. In Q4 2022, Data team's idea of leveraging users' dining behavior data aligned with Campaign squad's concept of telling the story behind users' dining journey. The project quickly took off with multiple teams collaborating in unison.

It was our first time implementing a "Year-in-Review" concept with limited resources, so the Campaign squad and Data team had a meticulous approach to ensure thorough coordination.



To implementation


The Look 


Designing and implementing the User Interface (UI) presented additional challenges for our Product Design (PD) and Front-End (FE) teams. Unlike previous campaigns that primarily emphasized graphic elements rather than data and animation, the 2022 Digital Lookback campaign marked a major change.


a. The screen layout must be flexible to display any given dish names and Merchant names with varying lengths. Also it must have room for graphic elements and entrance animation.

b. To ensure smooth animation between pages, PD had to elaborate on all animations in the prototype so the FE team could deliver it exactly how it was designed.


The Data 

To work efficiently within a short time, the Data and BE teams followed these simplified steps:


1. Defining the schema

Data and BE first agreed on a logical data structure that met the content requirements.


2. Exporting data according to the schema


Upon the agreed schema, the Data team analyzed various data types that they needed to collect and process. 


a. Some data types can be extracted directly from historical data, such as top dishes or Merchants.

b. Other data types required additional processing, such as food category level information. To do this, order data was processed by a classification model. For example, dishes like "chicken rice with crispy skin", "garlic chicken rice", and "Hainanese chicken rice" were automatically classified under the "Chicken Rice" group. After that, a humorous copy was generated accordingly: "Thank you for growing the Chicken Rice industry." 

After processing, the data was transferred to the BE team under .csv format instead of API. Mostly due to time constraints, but also to save development resources.


3. Integrating data into the system


The .csv format is not the best in terms of performance and efficiency, so the BE team must calculate the data storage, estimate the incoming traffic to make sure that users can access the campaign as fast as possible. With those estimates, the necessary infrastructure was in place by courtesy of the superheroic Site Reliability Engineering (SRE) team.


4. Designing APIs for client device access


Finally, BE designed the APIs for client device to access to the campaign page: 

a. One API to return the "static" parts showing historical data,

b. Another API to return the "dynamic" pages that encouraged user interaction, such as favoriting a Merchant on the "Promote the Development of the Food Industry" page.


With this campaign, it is clear that when properly executed, personalized campaigns will increase brand love because they hit the right insights and deliver a great user experience. This also motivates us to develop more personalization features in the future. Stay tuned!


*Data measured by BAEMIN Vietnam’s tracking tools.


Written by: Quang Nguyễn. - Data Scientist, Niki - Product Designer, Cong Nguyen - Back-End Designer 
Edited by: Thuyên Vũ - Developer Relations Associate

Share on: