UX and System Architecture

By Thane on Tue 12 July 2016

When building software systems, what exactly is the relationship between user experience (UX) and system architecture? Or, I suppose a better formulation of the question would be: what sort of relationship between UX and system architecture provides the best possible products for users? (Assuming, of course, you're addressing a user need in the first place ).

In Theory

There are a couple of different angles in trying to answer this question. Firstly, there's the theoretical perspective. For this, I'm going to refer back to the Mythical Man-Month (p. 45):

By the architecture of the system, I mean the complete and detailed specification of the user interface.

This is quite a different spin on my current connotations relating to "system architecture". It's far more user-centric than the various approaches I've seen in practice in the organisations in which I've worked. The majority of the time it's highly focused on the technology stack and how the various different interfaces fit together. Of course, this potentially has the long-term effect of providing for good UX, but the technology focus often doesn't provide explicit end-user goals to focus developers' effort. Often the goals end up being rather functional, with end user experience suffering as a result.

In the worst cases, as often seems to happen in practice, UX problems in systems often only crop up long after the product has actually been implemented and is running out in the wild. Here, it's often far more difficult and painful to restructure the system to improve users' experience.

The Intuition

There's also the intuitive perspective: intuitively, I would expect that starting with very specific user-oriented goals would provide a great scaffolding for developers to focus and guide their efforts. As per this video from some members of the Android development team, great UX involves function and form working well together holistically, in ways that afford the user a great experience every step of the way.

Just as an example, let's say you're building a web-based and RESTful service comprised of many microservices and different technologies. Apart from the visual design goals, some of your system goals could be things like:

  • End user page/API query load times need to be kept under 2 seconds. This has a massive impact on the way in which you structure your microservices, where you'll put load balancers, where, geographically, you'll set up CDNs, what kinds of databases you'll choose, etc.
  • Must facilitate offline access as far as possible. If, for example, you anticipate that your users will be accessing your service from areas where connectivity isn't always available (as is often the case here in Africa), your technical decisions will be heavily influenced. You'll need to think about using local storage in mobile browsers, as well as caching on mobile devices for API queries, with as much functionality as possible built into the client portion of your application, and as little as possible built into the server-side portion.
  • Scrolling through long lists must be fluid - no jerky scrolling, and load items (when connectivity is available) must be invisible to the user. This not only affects the client portion of your application, but also the server side. Depending on what sort of connectivity your users have, you may need to adapt the quantity of data you request from the server to facilitate a fluid experience for the user. In the aggregate, this has a massive impact on the server load, as well as the manner in which the client caches data, and how much data the client will end up caching.

Great for Testing

Not only would setting explicit UX goals for a product before its construction seem to provide focus for developers and guide effort, it would also provide really practical, hands-on testing criteria that pretty much anybody in the team can validate. It's also relatively easy to see how certain kinds of tests (like the page load/API query times) could easily be automated, even though some kinds of tests are very subjective.

(The fortunate and unfortunate reality of UX is that we'll never really be able to get away from subjective testing completely, because the reality of products is that they're ultimately built for people, who are, by their very nature, subjective beings - products for products' sake, or systems for systems' sake, tend to not get very far in terms of user uptake).


It makes sense then that the "systems architect" (if that role is a defined role in your context, as it is in mine) also be a champion for the UX of the product. In fact, if we stick more strictly to Brooks' definition of system architecture, the UX practitioner is, in reality, the actual architect of the overall system.

The challenge here is that it seems as though we've already built up an entire culture around the concept of "systems architect", where their role involves very little practical focus on UX, and the UX role is separated out almost completely. Not only is the UX role separated out, the UX practitioner is often only considered after the technical solution has been designed, and is often only associated with the visual aesthetics of the final solution (a legacy perception stemming, I think, from old, bad web design practices).

This is something that requires a culture change in such organisations, which, as we all know, is really, really hard if it's not baked in from the beginning of a group or team's inception.

Your Experience

Have you seen more user-centric approaches to system architecture in your own organisation(s)? Let me know what you've seen working in practice!