AutoCAD Accuracy and Touch Input


AutoCAD is a platform used by architects to draw their plans. Their main need is the ability to input exact values to geometries in their drawings.

The transition from mouse and keyboard to touch-controlled interface created the following problems:

  • Users struggled to find the keypad
  • The mental model that users formed was different than the way the system works
  • Minor usability issues lead to high error rate and decrease in efficiency

As a product designer, my goal was to further define the touch experience in terms of performance  and also make it efficient and useful.

Product autoCAD mobile app
Role  Product designer
Industry Computer-Aided Design
Year 2019


Keypad awareness

Problem: Users struggle to find the keypad

Even though we added a keyboard to the product, users continue to complain about its absence and only ~35% of the users evoked it at least once.

Exploration: how might we improve affordance

After competitor analysis and brainstorming, we came up with several optional concepts which were prioritized according to the estimation of effort and impact.

The concepts we explored:

Add icon

FAB button

Bottom sticky-panel

Onboard with coach marks

Solution: small signifier, great impact

We decided to test the version with an icon to increase affordance and hint users because it had the lowest effort estimation. To test it we conducted a/b testing between one version with the keypad and without it.

The concepts we explored


New Interaction Model

Problem: Wrong mental model

Due to a technical limitation, the keypad allowed entering values for the NEXT segment and not for the CURRENT one as AutoCAD users are used. This resulted in high error rate and in observations we saw that users failed to accomplish the most basic tasks

Solution: touch "mouse" interaction

We designed an interaction model that imitates the mice interaction.

  • Drag gestures move the crosshair
  • Tap gesture acquires a point in the crosshair's location


Cursor state feedback

Problem: Users Don't understand this interaction

Observations surfaced a low learning curve and some reoccurring errors.

Users used the gestures with some expectations about the result was either different than expected, in other cases nothing happened after users' actions

Solution: System state feedback

In order to inform users about the current system state and what can be done, we added feedback to the crosshair to inform the user when a tap is acceptable when a tap is invalid and instructions to recover in case of a mistake


Holistic keypad

The next goal was to improve performance and create a delightful experience while entering exact values

Problems: minor usability issues

Usability evaluation - by usability sessions and heuristic analysis - we found the following usability issues

  • Users where not able to predict the outcome because they had no live preview
  • No easy recovery from errors. only UNDO or DELETE and repeating the command
  • Lack in sense of control because they keyboard hide most of the screen
  • The layout of the native keyboard is not efficient for entering numeric values
  • Users are using invalid charachters
  • Hand gazing from moving the hand between geometry to the keyboard


A custom inline keypad to appear next to the geometry and prevent hand gazing.

It will have all the special characters to each unit type and allow live preview so the user will be able to predict the outcome of their actions in order to feel in control and avoid errors

The layout adjusts itself according to the file drawing units




Surveyor's units




The overall results of the above showed improvement to some important metrics



keypad awarness



Adoption rate


of all point acquired

without mistakes

Keypad Awareness

Increase of


In A/B testing, the new icon version showed an increase of ~30% to the proportion of users that uses the keypad in comparison to users that didn't have the icon.

Adoption Rate

Increase of


In percentage of new users that opened at least 4 drawing in their first month after adopting the new interaction model

New interaction model

After adding the cursor state feedback alongside with the new interaction model the percentage of acquired points without mistake by all app users increased from 91% to 96%


Of the users that made a mistake in the first acquired point managed to acquire the second point without any mistake within 8 seconds