Problem
•
Need to fetch the boost id of every boosts used by the user from the boost table when you apply certain criteria (recent purchases, frequent purchases) to recommend boosts suitable for the user.
◦
This work based on the user is repeated for each recommendation.
◦
The heavier the user, the higher the expected work cost.
•
To make recommendations based on the criteria of “boosts recently viewed by user”, you can utilize the data currently accumulating in Amplitude’s data lake. - Reference
◦
As this has never been done before, you must consult with the data team first.
Actions
Boost recommendation cost (if speed improvement is needed)
1.
Get help from the data team.
a.
We can perform the calculation first from the data team’s service, and then send it to API.
2.
Slightly modify the recommendation criteria.
a.
Example: Only include details within 30 days in the criteria
Select #2 based on v1.0
Boosts recently viewed by user
1.
Whenever a user views a specific boost, send that information to a separate backend API endpoint.
a.
Create a table with a unique row for each user, and update the table to maintain a certain number of recently viewed boosts for each view attempt.
2.
Get help from the data team.
a.
Data is already accumulating in Amplitude, so find a way to utilize the data.
