Usability Concerns for AR Apps

Blog / Stan Kowalczyk / November 23, 2016

After years of speculation and hype, augmented reality (AR) apps are finally gaining traction in the consumer space. By overlaying content over the top of what our phone cameras display, AR apps allow us to add things that aren’t present in reality or replace parts of the real world with the virtual world.

We see AR in our games, in our home improvement apps, and even in some language translation apps. In the first case, AR helps bring our virtual games into the real world, for home improvement this means that we can see what our existing kitchen may look like with a renovation, and for language translation apps it means that we can read signs, shopfronts and posters in our native tongue. As AR apps gain popularity we need to help our users get the best experience and make the most of our apps.

Our Experience

At DSTIL we recently worked on a project involving the development of a prototype AR app. Thanks to this app, potential customers will be able to see what their bike may look like with modifications or accessories applied to it. By showing the approximate end result, customers will able to more easily realise the consequences of modifications that can permanently alter their bike.

Fig.1 – Photo of bicycle.

In order to show the customer what their bike may look like, modifications and accessories need to look and feel as natural and indistinguishable as the rest of the bike. This means that a high degree of accuracy is required when overlaying a new seat, wheels, or any other accessory – each one has a specific place to go on a bike. To aid in achieving such a result during the prototype phase, the angles from which a bicycle could be photographed were limited.

However this posed an interesting problem – if you apply constraints on the angles of taking the photo, would users be able to use the app effectively? We tackled this problem by performing several usability tests.

Usability Tests

When conducting usability tests, our test candidates were only told about the purpose of the app, but not how to use it. In this way, we could observe and record their unbiased interactions with the app.

Our test candidates varied in skill level and experience, from those who self-identified as tech-savvy, to those who had used smartphones only as communications tool (i.e. for phone calls and text messages).

A behaviour that we immediately noticed (and anticipated) regarded the position in which a smartphone was held. Candidates always began the test by holding the smartphone as they would a digital still camera: at shoulder / chin height, and at a comfortable distance away from the eyes (generally at a distance of approximately 15 centimeters).

Fig.2 – Person taking photo on smartphone.

After some experimenting, most test candidates (without external influence) began trying different camera heights and angles relative to the picture subject.

Possible Solutions

The app used during the usability tests was not designed to help or assist the user in any way—this worked to our advantage. The lack of existing solutions allowed us to be unbiased in experimenting different and innovative solutions.

Two of these solutions can be applied to a broader range of AR apps.

The first (and easiest to implement) solution is in app education. This can be implemented in many ways; one way is to show the user some getting started text, tips, or video when the app is first opened, or a combination of all of them. The main issue with this solution is that users generally skip first-time tutorials, and many resources on this problem can be found online (e.g. here).

Fig.3 – Smartphone Tip Gif.

A second solution is providing the user with visual cues while using the app. You may have already experienced something to this effect if you’ve used the panorama feature on a camera, or turned on visual guides to help you avoid taking photos on an angle.

The implementation of visual cues is app-specific and can vary greatly in different use cases. You might perform some image processing and show the outlines of your subject matter within the camera field-of-view (e.g. edge detection might be applicable depending on what your app does). Moreover, dynamic visual cues based on positioning sensors (e.g. accelerometer, magnetometer) can be used to help correct the way your user holds their smartphone.

Fig.4 – Size of clusters generated by the k-medoids clustering algorithm applied to the commit messages of DSTIL’s Wicketkeeper project.


When developing your AR apps it’s important to consider any constraints you may have and what steps you can take to address them in order to minimise any impact they may have on the natural behaviour and responses of your users. Whether you use in app education, visual cues, or something else it’s important to think about how you can help your users get the most out of your app.

Header image courtesy of Stan Kowalczyk, Deakin Software and Technology Innovation Lab. Embedded photo of elephant courtesy of Alessandro Desantis.
Fig 1 – Courtesy of Alexander Shustov.
Fig 2 – Link.
Fig 3 – Courtesy of Stan Kowalczyk.

Thanks to Nicola Pastorello, Niroshinie Fernando, and Matt Hannah for proofreading and providing suggestions.