If it doesn't exist already, I want to propose a variant of Conway's Law which states

Any organization that designs a system (defined broadly) will produce a design whose structure is a copy of the organization's communication structure

The variant is

Any designed user experience will reflect the underlying data structures

An example - a video streaming sign up flow

Let's take a sign up flow, where the user is going to enter their details, as well as entering some of their TV watching genre preferences.

There are multiple ways we might design a data model for such a system:

Approach 1 - Include the preferences as part of the user object

type Genre = "horror" | "romance" | "action" | "comedy"; 


// GET, POST /user
type User = {
    username: string; 
    emailAddress: string; 

    favouriteGenres: Array<Genre>; 
}

With a data model like this, a user could enter all their details on the one page, and then submit the form, we POST /user, and we're done.

Or if we wanted, we could split the flow into two pages, having the user select their favourite genres after they've entered their details, but this just being a frontend facade, we don't actually submit the data until they've finished selecting the genres.

However, this approach does have a little more friction from a development perspective, there's that little bit of extra logic required to split the form across two pages and then join the data again for form submission.

Approach 2 - Genres are a collection belonging to users

type Genre = "horror" | "romance" | "action" | "comedy"; 


// GET, POST /user
type User = {
    username: string; 
    emailAddress: string; 
}

// GET, POST /user/{userId}/favouriteGenres
type FavouriteGenres = Array<Genre>; 

If our data model was way, the single page flow is less natural - we need the id of the user before we can call the POST /user/{userId/favouriteGenres endpoint.

We could fudge it, making it a single page, but hiding two API calls in the one form submission. But then we have issues like, 'what if the first API call succeeds, but the second one fails'?

We can see that a two page sign up has less friction, the user experience is going to be inclined this way, because of the data model.

What of it?

I don't know really. It's an interesting observation to mull on.

I'm not really suggesting that we should change our data model to facilitate the best user experience; instinctively, I would say that considerations around performance and the mental sensibility of data model are more important.

Questions? Comments? Criticisms? Get in the comments! 👇