We were required to design a household central air condition. Our target customers are Chinese families. This project was 9 weeks.


This is a cooperate project between Tsinghua Human Factor Lab and an international company. My team had 3 students and a supervisor. Each of us three needed to design a thermostat on different focus. One was designed for the elderly, one was for automated control, and mine was focused on way of input and output. Then, we did a 2-week user testing to compare all our three plus existing two methods.

My role in the whole project was designer of the following solution, one of the two conductors of user tests and the analyst of data.

Explore the Problem Space

Literature Review & Conpetitive Product Analysis

 – Literature Review –

Through researching on published papers, we identified following trends in thermostats:

  • Voice-controlled thermostat
  • Larger LCD
  • Zonal control
  • Receive price or emergency signals
  • Remote control
  • Response to signals, not simply temperature
  • Integration with other devices

 – Competitive Product Analysis-

We first analyzed two leading products: Nest and Honeywell Lyric.


Nest is controlled by rotating and press outer layers. We also researched its functionality:

  • Auto-schedule
  • Auto-away
  • Report Generation
  • Remote Control
  • Sunblock
  • Time to Temp
  • Airwave
  • Auto-tune
  • Cool to Dry

Lyric is controlled by rotating outer layers, press on the two buttons in the front face of it and a small touch screen. It has functionalities including:

  • Auto-away
  • Smart Cues
  • Remote Control
  • Motion Sense
  • Short-cut Setting
  • Fine Tune

To identify the facts and trends in China, we also analyzed famous brands inside China.

Chinese Companies

Define the Problem


 – Functionality –

Through comparing Chinese thermostats with Nest and Lyric, we found chinese thermostat fails in providing smart control/ schedule and energy/ spent saving functions. While they provide some key functions to Chinese customers, such as air quality monitering. We assume our product need to cover these functions. To validate our assumation, we send out questionnaires.

Through the result of our 72 questionaire, we found:

Smart Control/ Schedule

According to our questionnaire, most informants have regular air-conditioner using habit, which making “smart control/ schedule” possible.

Energy/Spent Saving
Save Energy-01

This function is preferred by many informants. As our target user usually owns large house, the saving would be a large amount of money.

Air Quality Monitoring

Due to the polution in big cities of China, this function acts as an edge for the fierce competance among similar products.

 – Output & Input –

We showed our informants six different types of thermostats, as showing in following pictures and ask them to sequence them according to their preference.


Here is the interpreted results:

Input Preference: (1) Informants likes to use knob to control temperature. (2) Informants dose not feel confident in controlling thermostat by directly expressing feelings. (3) Touch screen is welcomed.

Output Preference: (1) They prefer larger screen. (2) Their feeling towards graphic data visualization polarized.



Considering most respondents prefer round-shape traditional-thermostat-like design (knob control) and large screen, we would adopt design like Honeywell lyric, and Nest, while making screen larger.

To make APP coherent with controller on wall, the main page of APP would be designed like the main UI page of controller on wall.


Schedule Mode

Users can use the on-wall controller to control the temperatures and switch between “schedule” and “event”.

Every temperature choice they made will be stored and used to alter the schedule smartly.

Event Mode

To interrupt smart schedule to prepare for event at an unexpected time, users just a simple tap to change the mode to “event mode”.

Users can set different behavior for different events on mobile App and switch between events either through App or on-wall controller.


 User Testing

We had 50 subjects in our user tests. 25 of them were Chinese, and 25 of them were international students. Each of the subjects needed to finish three set of questionnaires and finish tasks on five controllers. We videoed the whole task performing process.

 – Experiment Process –

thermosta- test

 Data Analysis

The data to be analyzed mainly come from video data and three sets of questionnaires. We have three indicators for each design: workload, preference, and satisfactory level. The independent variables include (1) types of design, (2) platforms (one is on ipad while others on iphone), (3) automation, (4) area where subjects came from, (5) subjects’ expertise on smart thermostat, and (6) frequency of using App to control thermostat/ air-condition.

 – Workload –

We had three sub-indicators of workload represented in task performance video: path length, times of confusing and task completion time.

  • Significance Analysis: (1), (3), (4) is significant.
  • (1) Type of design: My design requires the second most workload. (2/5)
  • (3) Automation Level: The higher automation level, the higher workload.
  • (4) Area: Chinese users reported least workload and European users reported the highest workload.

Explain of Result: My design offered a lot of functions, which increased the workload. We also analysed the workload of my design on individual task:

According to NASA-TLX Scale, we divided the workload represented through questionnaire in six dimensions: Mental Demand, Physical Demand, Temporal Demand, Performance, Effort, and Frustration. To calculate the total points for every design solution, we asked subjects their attitudes on how much each dimensions mean to them, and we used that as a weight to calculate the total points. Use that as the dependent variable, area and types of design solution as factors, we drew a interactions plot as showing below:

thermostat-interactions plot

Takeaway: From the plot we could see the workloads reported by Chinese users (black line) on were the lowest on almost every single design solutions. The tolerance of Chinese users are high, which shows a big chance and call to design for them.

 – Satisfactory –

We had four sub-indicators for satisfactory: the perceived usefulness, easy to use, the attitude towards using, and the willingness to use it again.

My design rated 1/5 in perceived usefulness, 3/5 in easy to use, 2/5 in attitude towards using, and 2/5 in willingness to use it again in Chinese participants.

Takeaways: When we drew the graphs out. We found people from different areas have different feelings towards different designs, but people from other countries in Southeast Asia (red line in plot) reported similar feelings as Chinese users (black line in plot). Thus, if we designed a thermostat for Chinese users, hopefully, we could also promote this product in other countries of Southeast Asia.


 – Preference –

We collected Chinese and Non-Chinese’s preference towards different designs.

Most Favorite (Chinese)

  • Design One
  • Design Two
  • Design Three (Mine)
  • Design Four
  • Design Five
  • Design Zero (Physical Remote Controller)

Least Favorite (Chinese)

  • Design One
  • Design Two
  • Design Three (Mine)
  • Design Four
  • Design Five
  • Design Zero (Physical Remote Controller)

Most Favorite (Non-Chinese)

  • Design One
  • Design Two
  • Design Three (Mine)
  • Design Four
  • Design Five
  • Design Zero (Physical Remote Controller)

Least Favorite (Non-Chinese)

  • Design One
  • Design Two
  • Design Three (Mine)
  • Design Four
  • Design Five
  • Design Zero (Physical Remote Controller)

From the Pie Chart we could tell, the most favorite product by Chinese subjects is my design. It shows the similar result as in Satisfactory analysis.

Although design three does not perform best in “least favorite by Chinese” or any result in non-Chinese, the results in these three parts are acceptable. We were not designing a product to satisfy everyone. We just need to make sure a promising part of our target user will pick our product their favorite product.