Why LOVETRIP Didn't Send External Responses Straight to the Screen
Converting travel data from several sources into an internal place model, and how separating collection from service helps absorb external changes.
Joohwan Yu · Founder, Full Stack Developer
Toris · Published
Contents
LOVETRIP was a product that tried to gather scattered travel information into a one-screen planning flow. To users it looks like places and courses, but internally I had to handle data from several sources. Even the fields describing the same place differed by source, and the refresh timing wasn’t the same.
The Moment an External Format Becomes an Internal Rule
Passing external API responses straight to components is fast at first. But once each screen starts handling missing values and different field names, the external format spreads all over the code. When one source changes, you may have to edit the recommendation logic, the map, and the detail screen together.
In LOVETRIP, I chose to handle external data by passing it through a conversion boundary into an internal model. I also separated the collection job and the service that handles user requests as distinct execution units. The thinking was to keep a collection failure from leading directly to a failure of a user request.
Conversion Was More Than Renaming
Copying an address string over is different from checking that a location is valid. Whether to fill in a default when a value is missing or exclude the item from recommendation candidates is also a product decision. If I rebuilt it now, I would make the source, the refresh time, and the reason for any conversion failure traceable in the model and the operational records.
In particular, “no information” and “free of charge” must not be expressed as the same value. If you put 0 because you don’t know the price, budget recommendations can look more optimistic than reality. If the conversion layer erases this kind of uncertainty, the screen is clean but the judgment is inaccurate.
Why I Made the Data Source Visible
My earlier write-up also recorded the background for showing the data source. When several kinds of information are mixed, users wonder whether it is current. Letting them check the original, rather than just showing the recommendation result, was needed to keep trust. Still, the mere existence of a source link doesn’t guarantee current opening hours or prices.
This experience made me see fetching external data and making it usable in the product as separate jobs. Connecting the API was the start. Deciding what the data means, recording what you don’t know as unknown, and handling changes at a single boundary were the center of real maintenance.
Related post: LOVETRIP Design and Trade-offs
Advertisement