Spring financial - Personal loan application flow
Optimizing API Search Call Efficiency for Cost Reduction
I worked on a project for Spring financial personal loan application, optimizing their API search calls to reduce costs and improve efficiency. The goal was to refine the logic for data retrieval by analyzing user input behaviour and determining the ideal timing for API calls. I designed a solution to minimize unnecessary requests by identifying the most efficient intervals (in milliseconds) to pull data based on user interactions. This reduced redundant API calls while maintaining a seamless user experience and lowering operational costs.
Understand the logic, improve result accuracy.
Steps:
Test various address types.
Identify search keywords for each type.
Calculate the average time to find the correct result.
Search behaviour Testing & Search Keywords
Findings
Users need to type the full address with the unit number to start, as we don't provide the option to edit at the end. The first 3-4 digits only count as the unit number if applicable.
Users won't know what to type to get the best result until they fail on the first attempt.(search with street number.)
If the address with 'unit #' is not in the system, the user won't be able to find their address; they need to type it in manually.
Users need to provide at least one or two characters of their street name to find their address on the list (e.g., 1644 N, 2038 cl ...).
When the address contains information like 'ave west' or 'ave east,' users won't be able to search in the normal order; they need to add the condition 'w' or 'e' at the end of 'ave.' e.g. 2080 2nd Ave E VANCOUVER BC V5N 1E7 CANADA
Response time metrics
In 500ms: If the user types their address without a pause or with a slight pause between the number and street name, they will most likely receive 1-2 results (closer to 2 call).
In 700ms: If the user types their address without a pause, they will most likely receive 1 result. Users need to provide the street number and 1-2 characters of the street name to find the correct result.
Design Solutions
Display suggested keywords to help users get accurate results quickly.
Show a supportive message when the keyboard is active.
Add a "Type manually" option in all flows to avoid repetitive searches.
Guiding Users to Search with the Right Keywords-Supportive message
Since Canada Post's search system has a unique way of handling address searches, providing clear cues can help users find accurate results more quickly, leading them to their goal faster.
Display supportive messages as captions that users can easily access at any time for guidance.
Provide a secondary guide for users with complex addresses.
Provide a "Click here to type manually" option to prevent users from making multiple search attempts.
If we pull the data after 700ms, users will likely be able to type up to 4 letters for better results. Additionally, provide a "More" button for the same address with multiple units, guiding users to select a unit and reducing search time.
Users can always access the "Click here to type manually" option, reducing frustration and preventing multiple attempts.
