September, 2014

UI Treadmills: An Inline Scroll Problem

As a user scrolls down the page using either a mouse wheel, trackpad or touch screen, they can become trapped by an inline scroll. Whilst trapped, they are still scrolling, but the page that they were scrolling before now becomes stationary. This situation can be likened to using a Treadmill, where you are moving, but your movement does not have any result in your position within the broader environment. In these situations, it is highly probable that the user's intention was not to scroll this subset of the page.

In order to exit the 'Treadmill', the user must either move their cursor or finger outside of the area the inline scroll occupies, or wait until they reach the bottom of that content, with scroll focus then returning to the main page. To "step off of" the Treadmill, so to speak.

How Inline Scrolls Come About

Content, such as text, occupies space on a page. If you constrain that content to a size smaller than it needs, then you are left with an overflow. Much like trying to pour a pint of water into a teacup. Eventually, it will overflow. In software, you are left with two options with this overflow when constraining; cut off the content, so the overflow is not seen (not really an option) or to allow scrollbars (potentially dangerous grounds!).

HTML elements that typically cause this are either 'iframes', or 'div' with a fixed height or width attribute value. The rational behind their use being to display content from an external site or system in the first case, or to minimise the presence of the content the div encapsulates, on the page.

What is The Problem?

Cursor or finger position on the page acts as a cue to the computer as to what the user would like to scroll. As the user travels down the page, what is below their pointer changes. This means that a user can inadvertently enter a region of the screen that has its own scrollbar, and then they would be scrolling that area instead of the containing screen. In the video below, there are three different examples of how cursor placement impacts the way the user scrolls the page:

What were the three errors when scrolling?

1. Cursor was not placed over a scrollable region
The location where the cursor is first placed in the video is not scrollable. Once the user eventually manages to scroll down the page, you will notice that the global navigation, actions bar, and heading remains in the viewport. In this case, they used a scrollable div element to make this, but instead they could have used CSS to achieve this "Sticky UI".

2. Cursor was trapped within an inline scroll
Once the cursor was placed in the middle of the screen, it began scrolling the scrollable region. Here the user got caught within the posts feed, but continued scrolling and eventually got to the end of that feed and scroll focus was given back to the parent container.

3. Cursor was trapped on the map
Finally, the cursor went down the left column, and ended up getting trapped within the map. In this situation, no matter how much the user scrolls up or down, they will never exit the map; they must first place their pointer outside of the map and then scroll, in order to continue scrolling the page.

This particular case study is a pungent one. When you begin to focus on where the user can scroll, you start to realise that there is not many locations on the page that they can safely place their cursor and travel down the page unhindered.

The areas highlighted in green allow the user to scroll unhindered.

The areas highlighted in green allow the user to scroll unhindered.

In the annotated screenshot above, the areas highlighted in green are areas the pointer can be placed without ending up on a 'Treadmill'. This means only 22% of the viewport is safe for a cursor to reside in. Should the cursor be placed in the remaining 78%, then the user will have to eventually move their cursor to one of the 'safe zones', which are not clearly marked like my annotated screenshot; they are not guaranteed to find the 'safe zone' in the first or second attempt either.

Moving a cursor to an exact location is not a quick action, as Fitts' Law will attest to. This all makes for a feeling of a loss of control, fiddly, and frustrating experience.

Remedying Treadmills

There are a number of design directions you can take to avoid or minimise the impact inline scrolling can have on an experience. Can we display on the page without the need of an additional scrollbar? Is the content list-like in nature, so that it can be paginated? Can we display it on request using progressive disclosure or a thumbnail?

In the Dynamics CRM 2013 example, the persistent header could have been solved by using a technique known as Sticky UI, which is a set of CSS rules. The posts feed could have either been paginated like the contacts section on the page, or had its own column and not introduce an additional scrollbar. Lastly, the map could have been an image, which when clicked, brings up a larger map view which is interactive; I would expect this would better fit the user's needs too, regardless of the Treadmill issue.

You cannot always avoid inline scroll. Many times, often to my displeasure, there has been a technical or business constraint which meant inline scrolling had to be used. But there are also common design patterns, like Master/Detail, which can make good use of this capability. In those cases, my advice would be to consider:

  1. Minimising the amount of screen width it or they occupies
  2. Use styling to differentiate that area, should it be a border, or background colour
  3. Do not nest inline scrolls within each other

Finally, as with everything you are doing, test it! In Axure you can make scrollable regions by placing content within a Dynamic Panel, and then:

Right click on the dynamic panel > Scrollbars > Show Vertical Scrollbar as Needed.

There is an option for horizontal scrolling too, but that should be avoided, especially if you are doing vertical scrolling as well.

August, 2014

Avoid ‘Sinkholes’ by Using Closable Panels Instead of Accordions

Accordions suffer from a phenomenon I have coined as 'Sinkhole', which in its worst cases can cause both confusion and disorientation to the user. Sinkhole is however avoidable by using  Closable Panels, a pattern that is very similar looking, but behaves differently to Accordions.

What is an Accordion?

The Accordion component is made up of two distinct elements; the panel title and panel content. This pair of elements are repeated as needed to house different sets of information.

Accordion example - Sinkhole

Only one instance of the panel content is viewable at a time, and a user switches between content views by invoking the panel title. It is this event of changing content views that brings about the Sinkhole phenomenon.


When an Accordion grows in height, the content below it is pushed downwards. Should an Accordion's height shrink, then elements below shift upwards to occupy the newly vacated space. An Accordion can grow or shrink each time the user invokes the Accordion's Panel Titles. This interaction brings about the phenomenon I have coined as a Sinkhole.

Microsoft's website uses a Mega Menu on large screen devices for its global navigation, and changes it into an Accordion for smaller screens. The small screen version of their navigation suffers from Sinkhole, exhibiting the undesirable traits I described, which is demonstrated below:

In the video the user first opened 'Products', and after scrolling down the product list, they then chose to look at 'Downloads'. When 'Downloads' was tapped it caused 'Products' to collapse, which shrank the height of the Accordion substantially. At that moment in time (8 seconds passed) the user can no longer see the navigation menu. It has disappeared from view, as it is now above their viewport, and this can cause confusion and disorientation to the user.

Remedying Sinkhole

Sinkhole, what the video above just demonstrated, is quickly remedied by switching from an Accordion to Closable Panels. Then panels are independent from each other; they only close when a user clicks/taps it again (a toggle), and opened panels remain open should the user choose to open another panel. The opposite of what an Accordion does.

In order to close a panel, the user must first scroll up to see the panel title, so that they are able to click/tap it. The page content then flows upwards towards the user's viewport, which means the user's viewport is not changed by the page reflow, keeping them where they where before invoking the panel title.

Inline Scroll

An Accordion's height can be fixed using CSS. When the Accordion's content becomes too much for it to display, a scrollbar is provided so that all of the content can be viewed. The trouble then is how big to make it? Too small when displaying lots of content and it becomes troublesome to consume that content. Too large, and it starts taking up more page space than necessary, which contradicts its very purpose of existence.

Inline scroll has its very own phenomenon too, which I have called 'Treadmilling'. This happens when you scroll down a page using a mouse-wheel or trackpad gesture and your cursor passes over the area where the inline scroll is. Then your scroll begins to scroll through that content instead of the page, with you moving but the main page remaining still. This is a topic I plan to cover shortly, so I won't go into this any further here.

Considering this, I personally cannot find an argument against using Closable Panels instead of an Accordion, with or without inline scroll. Please let me know your thoughts using the comment section below!

July, 2014

6 Tips to Make Applications Feel Faster

Speed is an important factor, and if it is not accounted for properly, can make or break an application or service. Often the focus is on measuring system performance; identifying where the most time is being spent, and then optimising the offending area.

Complementary to this, there are several strategies you can employ to make an application feel faster than it actually is. What the user sees on their screen shapes their perception on how fast they think your application/service is.

Why Faster?

It is well documented that users do not like slow websites, or software for that matter. In business terms that means:

  • Amazon: 100ms delay results in 1% sales loss
    (potential $191m lost in revenue in 2008)
  • Google: 500ms delay drops search traffic by 20%
    The cost of slower performance increases over time
  • Yahoo: 400ms delay results in 5-9% drop in full-page traffic
  • Bing: 1s delay results in 4% drop in revenue
  • AOL: Fastest 10% of users stay 50% longer than slowest 10%

- Stats taken from How to Make Apps Feel Faster by Luke Wroblewski

Next we will cover different techniques to make applications feel faster to the end user.

1. Progress Indicators

Users need reassurance that the system is dealing with their request, and that it has not frozen or is waiting for information from them. Progress indicators are used for this very purpose, to signify it is working, and to set expectations as to when it will be complete and ready for use again.

The style of a progress indicator can influence the perception of speed, so much so that it can appear to be 11% faster. These results are achieved by applying ‘a backwards moving and decelerating ribbed’ treatment.

You can view a video of this research from New Scientist below [requires Flash player]:

You can also read the paper about this research

2. Optimistically Perform Actions

You can allow a user to be more effective with their time, and get the impression that an application or service is fast by not requiring them to wait for actions to take place. Thereby freeing them, allowing them to move onto the next action they need to take.

Instagram for example, begins the upload of an image early; once the user has passed the filter stage. Even though they have not yet added a caption, location, or even committed to the posting the image.

Instagram performing actions optimistically. Taken from Secrets to Lightning Fast Mobile Design by Mike Krieger

Instagram performing actions optimistically.
Taken from Secrets to Lightning Fast Mobile Design by Mike Krieger

Once the user has finished the upload flow, the image appears in the user’s feed, even if the upload is still in progress. It just happens to be a local copy of the image. But to them it appears as though it's already on Instagram's service.

Instagram reaps three benefits by making the image upload flow optimistic:

  1. Starting the image upload early in the flow gives them a head start, meaning it will be available to other users, sooner.
  2. Showing the uploaded image in their feed, even though it may not be uploaded yet, gives the user task closure by appearing finished.
  3. Users think their service is quick, even though typically uploading images, especially from a mobile, can be a slow process.

For the second point, the same applies to commenting on Instagram. Once you submit your comment, it appears beneath the picture immediately. Yet, that update is not actually instantaneous, it just appears as though it is.

Mike Krieger, co-founder of Instagram, goes into more detail about how Instagram makes their app feel ‘lightning fast’ in his presentation.

3. Distract

"A watched pot never boils"

Sometimes it is unavoidable, and an application has to become temporarily unavailable for a user, whilst it works on their request.

One of Bruce Tognazzini’s principals for interaction design states:

Offer engaging text messages to [keep] users informed and entertained while they are waiting for long processes, such as server saves, to be completed.

- Latency Reduction in First Principles of Interaction Design by Bruce “Tog” Tognazzini

A prime example of this is on the simulation game Football Manager, which regularly involves long processes:

Football Manager - Loading Distraction

Hints and tips are displayed whilst the users wait for results to be generated, or for games to be created. Messages are used to bring attention to new features, or educate the user so that they can become more effective at playing the game.

Another example is the web-based version of Balsamiq, a wireframing software, which shows the user quotes whilst it is loading:


When providing distractions to the user, it is important to bear in mind that they have a limited time to view the content shown to them. Once the process is completed, that information is taken away from their screen.

4. Progressive Rendering

When designing the pages of an application, the components you place onto a page belong to zones, such as a header or footer. Progressive rendering sends those zones to the user in a prioritised order.

Priority of the page zones are determined by the following factors:

  1. Page placement. Is it near the top, is it ‘below the fold*’ of the screen?
  2. Importance to the user
  3. Is the page asset much slower to return?

*In this case when speaking below the fold, the only reason I am mentioning this is because it is not presently visible, therefore less important compared to content above the fold.

Below is an example of how different zones of a page may be prioritised for an e-commerce website:

Wireframe of a fictitious e-commerce website. Page zones are numbered, giving an example how progressive rendering prioritises different parts of the page over each other. For example, a header is more important than a footer

Page zones are numbered, giving an example how progressive rendering prioritises different parts of the page over each other. For example, a header is more important than a footer

Progressive rendering gets key information back to the user quicker , rather than them having to wait for the entire page to be ready. This was exemplified by a study conducted by UIE, which found:, rated slowest by our users, was actually the fastest site (average: 8 seconds)., rated as one of the fastest sites by users, was really the slowest (average: 36 seconds).

The Truth About Download Time by Christine Perfetti and Lori Landesman

In this example, Amazon appeared quicker to users because information was displayed to the user sooner. It prioritised what the user could see first, and what was most important to the user.

5. UI Skeleton

This technique is closely related to what was just covered, progressive rendering, and is also known as a ‘ghost screen’. The first thing it displays to the user is the page framework; rather like a blank template. This ‘blank template’ is what distinguishes it from plain progressive rendering. Polar, a polling service, uses this very technique:

Parts of the template filled in overtime, once the information is made available by the server.

Parts of the template filled in overtime, once the information is made available by the server.
Image modified from Mobile Design Details: Avoid The Spinner by Luke Wroblewski.

The left-most screen is made up of a set of placeholders. Those placeholders are then gradually filled once the content is made available to the user interface, as the screens to the right depict.

6. Acknowledging Clicks

When pressing the play button on a cassette player, very explicit feedback was given to the user. The button sank lower, and a click sound was heard due to its mechanical nature. This all happened within a very short period of time, due to it being mechanistic in nature.

In software design, the artefacts users interact with are virtual, they are pixels. Good design traits, such as acknowledging interaction must be programmed into those entities.

Acknowledge all button clicks by visual or aural feedback within 50 milliseconds.
First Principles of Interaction Design by Bruce Tognazzini

Tabs on the web are a key example where prompt visual feedback needs to come into play. Even if the tab's content is not yet present, ensure the click/tap is acknowledged within 50ms. You achieve this by styling the tab invoked to active and the previously tab to inactive. Should the content not yet be available, then display a progress indicator, or perhaps offer a distraction, should the request be notoriously long.

You can also apply styling to buttons to indicate something is happening. A good design practice is to also trap multiple clicks of the button, so that the request the button triggers only gets registered once. This will help with speed too.

Jakob Nielsen has written about the 3 different response time limits, which is well worth a read.

February, 2014

Spotify Controller Using Leap Motion and Tobii EyeX

Two representatives from Tobii visited Avanade, where I work. They gave a presentation on the background of Tobii, the way their technologies are leveraged, and how developers can make use of their technology through the EyeX SDK.

Later that day we had a brainstorming session, and consequently developed a proof of concept using Tobii EyeX and Leap Motion to control a Spotify player.

Trying out EyeX

During the day we were able to try out Tobii’s EyeX controller on Windows 8.1. We used Modern UI apps such as Bing Maps, Windows Store, and Twitter. Since these kind of apps have been designed with touch in mind, it benefitted Eye Interaction as the ‘hit target’, so to speak, was much larger compared to UI’s designed for a mouse pointer. Larger hit targets allowed for improved accuracy when invoking UI elements such as tiles. Eye Interaction was facilitated by holding down a key with a special binding. This key allowed us to switch between modes, such as zooming in or out, or panning across a map, for instance.

Designing for Eye Interaction

Tobii shared their principals on what to consider when designing eye for interaction:

  • Eyes are made for looking around
  • Eyes and hands work well together
  • Eyes are curious
  • Eye movements provide information

Read more about these principals over at Tobii's blog.

Using these principles, we began a whiteboard session to explore how we use our eyes when using computers. We agreed that our eyes are “passive”, and that the clues our eyes give should supplement another method of interaction.

NUI - Whiteboard session

Whiteboard session on how to leverage eye tracking and motion detection.

We grounded this theory based on a study by UIE, where they looked at how users found flyout menus and rollovers, and discovered:

“We found users follow a pattern: they decide what they are going to click on before they move the mouse.”
Users Decide First; Move Second by Erik Ojakaar, UIE

In keeping with our Natural User Interface (NUI) theme, we wanted to try and combine Tobii EyeX with another gestural technology. We were fortunate to have both Microsoft Kinect v2 and Leap Motion available to us, which gave us some interesting capabilities to try and combine.


The concept we developed that day was a Spotify controller using Tobii EyeX and Leap Motion.  EyeX detects when the user is looking at the Spotify icon in the task bar. Leap Motion provided an interface where a user can give hand gestures to control Spotify. Gestures recognised by Leap Motion would not be honoured unless the user was looking at the Spotify icon at the same time as performing the hand gesture. The proof of concept application supported the following gestures:

  • Poke to play or pause
  • Wave right to play next track
  • Wave left to play previous track
  • Circle clockwise to increase volume
  • Circle anticlockwise to decrease volume
Spotify controller prototype using Tobii EyeX and Leap Motion

Spotify controller prototype using Tobii EyeX and Leap Motion

Why Leap Motion Instead of Kinect v2?

We chose Leap Motion over Kinect for our Spotify controller for the following reasons:

  • The user needs to be beside the computer, as the EyeX controller has a limited range of view.
  • Leap Motion has a much smaller desktop footprint, which suits close range interaction.
  • Leap Motion specialises in hand gestures, detecting each finger and thumb.


Having the eye tracking and motion capture capabilities as separate pieces of hardware quickly clutters your workspace. The fact that they are separate also means it is not really suitable for a laptop, as it requires a desk, and makes moving from one place to another quite cumbersome.

Many computers, like the one we used for the prototype, have media keys. Those keys allow you to change the volume, skip or return to a previous track, play, and pause. In terms of interaction speed, those media keys, although not formally measured, appeared to be considerably faster than using gestures.

Nevertheless, that day was a very thought-provoking experience. The capabilities on show were very impressive, and it will be interesting to see how they develop and are leveraged in the future.

©2021 Craig Woollard