A front-door lock can be a device, a security control or part of a leaving-home routine. Which relationships matter when people try to find and use it? We analyzed 36 open card sorts of 20 everyday smart-home activities. Security and entry produced the strongest agreement, while different category labels often accompanied similar overall card placements. Here's how those findings can inform the next navigation decision.
Why a connected home is hard to organize
In 2022, Google's Home product team said there was "no one size fits all model to organize the home." Its redesign brought together personalized Favorites, device-category Spaces and a separate Automations tab. Google's 2025 redesign organized the app around Home, Activity and Automations, while adding Ask Home for natural-language access. Even as the interface changes, the product has to support different ways to reach a function or understand what's happening at home.
A 2023 smart-home interface study used card sorting to group IoT devices before developing and evaluating an app prototype. A garage door might be considered an entry point, a security concern or one thing to check before leaving. The design challenge is deciding which relationships deserve a primary place and which need another route in.
A previous UXbeam study of government services raised a broader research question: can different card-placement patterns be understood through the organizing principles people use? Smart-home activities bring a different set of plausible relationships, among devices, household functions, places and routines. Here we look closely at what happened when respondents freely organized 20 familiar activities, including checking a window, letting in a guest and watering the garden.
Our analysis examined two parts of the results: which cards respondents placed together and the category labels they wrote. Fincher and Tenenberg (2005) distinguish these two ways of reading a card sort.
For the wider research approach, see our organizing-principles guide and the earlier government-services case, where two detected mental models highlighted a topic-versus-task distinction. The Smart Home results tell a different story.
See all 20 cards
Turn on the living room lights
Lock the front door
Check whether a window is open
Adjust the bedroom temperature
See who is at the front door
Start the robot vacuum
Turn off devices when leaving home
Set the house for bedtime
Prepare the house for your arrival
Reduce energy use while nobody is home
Check indoor air quality
Get alerted if smoke is detected
See which devices are using the most electricity
Let a guest into the house
Check whether the garage door is closed
Water the garden
Play music in the kitchen
Create a morning routine
Manage your child's bedroom devices
Check the home while you are away
Where the sorts agreed
Agreement concentrated in one part of the home. Of the 36 included sorts, 28 put "Check whether a window is open" and "Check whether the garage door is closed" in the same group, and 25 grouped the garage door with "Lock the front door". Several of the next-highest pairs also involved home security and entry, such as the garage door with "See who is at the front door" (23 sorts).
Those pairs are the exception. No pair of cards was grouped together in more than 28 of the 36 sorts, and 164 of the 190 possible pairs shared a group in fewer than half of them.
Cards placed together
Co-grouping count: the number of sorts that placed two cards in the same category.
Window open + Garage door closed28 of 36
Garage door closed + Lock front door25
Check home while away + Smoke alert23
Garage door closed + See who is at door23
Lock front door + See who is at door22
Reduce energy while nobody home + Turn off devices when leaving22
The list shows the highest-count pairs; the complete matrix is available below.
Darker cells indicate higher co-grouping counts. Both axes show the same 20 cards.
Accessible table: times each pair was grouped together
Times each pair was grouped together (36 sorts)
Card
Check whether a window is open
Check whether the garage door is closed
Lock the front door
See who is at the front door
Let a guest into the house
Prepare the house for your arrival
Turn on the living room lights
Play music in the kitchen
Create a morning routine
Set the house for bedtime
Start the robot vacuum
Water the garden
Adjust the bedroom temperature
Reduce energy use while nobody is home
Turn off devices when leaving home
See which devices are using the most electricity
Check indoor air quality
Get alerted if smoke is detected
Check the home while you are away
Manage your child's bedroom devices
Check whether a window is open
28
20
18
12
6
6
3
1
11
7
7
10
9
10
8
12
15
19
9
Check whether the garage door is closed
28
25
23
14
5
5
4
4
13
8
6
6
4
6
6
14
19
20
11
Lock the front door
20
25
22
20
9
8
6
7
17
8
7
7
5
9
2
8
16
19
12
See who is at the front door
18
23
22
21
10
5
6
4
11
7
5
4
5
7
6
8
17
20
10
Let a guest into the house
12
14
20
21
16
10
10
9
15
9
13
10
7
7
5
8
11
14
10
Prepare the house for your arrival
6
5
9
10
16
21
13
16
20
15
12
13
16
16
9
10
7
11
8
Turn on the living room lights
6
5
8
5
10
21
21
16
15
18
14
17
14
15
9
12
5
9
12
Play music in the kitchen
3
4
6
6
10
13
21
18
13
16
16
12
6
5
2
10
2
5
12
Create a morning routine
1
4
7
4
9
16
16
18
21
17
19
10
7
9
5
9
6
6
11
Set the house for bedtime
11
13
17
11
15
20
15
13
21
16
10
11
9
10
5
10
8
11
15
Start the robot vacuum
7
8
8
7
9
15
18
16
17
16
21
15
12
12
10
12
6
7
13
Water the garden
7
6
7
5
13
12
14
16
19
10
21
11
10
4
7
8
2
5
7
Adjust the bedroom temperature
10
6
7
4
10
13
17
12
10
11
15
11
18
17
15
15
2
3
14
Reduce energy use while nobody is home
9
4
5
5
7
16
14
6
7
9
12
10
18
22
21
13
9
11
9
Turn off devices when leaving home
10
6
9
7
7
16
15
5
9
10
12
4
17
22
19
8
9
10
12
See which devices are using the most electricity
8
6
2
6
5
9
9
2
5
5
10
7
15
21
19
17
15
9
10
Check indoor air quality
12
14
8
8
8
10
12
10
9
10
12
8
15
13
8
17
12
11
8
Get alerted if smoke is detected
15
19
16
17
11
7
5
2
6
8
6
2
2
9
9
15
12
23
14
Check the home while you are away
19
20
19
20
14
11
9
5
6
11
7
5
3
11
10
9
11
23
12
Manage your child's bedroom devices
9
11
12
10
10
8
12
12
11
15
13
7
14
9
12
10
8
14
12
Cards with split sorts
Forty-five of the 190 card pairs were grouped together in 40% to 60% of sorts. 'Manage your child's bedroom devices' was most often paired with 'Set the house for bedtime', in 15 of 36 sorts. That task could fit under rooms, household members or routines. The results alone do not tell us which organizing principle an individual respondent followed.
'Turn on the living room lights' was grouped with 'Prepare the house for your arrival' and 'Play music in the kitchen' in 21 sorts each, and with 'Start the robot vacuum' in 18. These are overlapping pair counts: one sort can contribute to several.
Main directions
Card placements
UXbeam Mental Model Detection (which groups sorts by how similarly they placed the cards) found one dominant detected mental model covering 30 of the 36 included sorts. Two small detected groups of two sorts each also appeared, and 2 sorts were not assigned to any.
The dominant detected mental model reflects similarities in card placements, not identical arrangements. This overall similarity can coexist with modest agreement on many individual card pairs. The two small detected groups are exploratory.
Category labels
Some category labels described household functions such as security or energy. We examined three other label types: action labels (7 sorts), frequency or necessity labels (5 sorts), and device-control labels (5 sorts). These label types appeared in 16 of the 36 analyzed sorts, with one sort counted in two rows.
The dominant detected mental model included all 5 frequency-or-necessity sorts and 6 of the 7 action-labelled sorts. Device-control labels showed a different distribution: 2 of 5 sorts belonged to the dominant model and 3 to the two small detected groups.
Label types and detected mental models
Action labels 7 of 36
Dominant detected mental model6
Small detected groups1
No detected mental model0
Frequency or necessity labels 5 of 36
Dominant detected mental model5
Small detected groups0
No detected mental model0
Device-control labels 5 of 36
Dominant detected mental model2
Small detected groups3
No detected mental model0
Each mark represents one sort. Label types come from our reading of the category labels; they cover 16 of 36 sorts, with one sort appearing in two rows. Detected mental models come from card placements.
Accessible table: label types and detected mental models
Label types and detected mental models (36 sorts)
Label type
Dominant detected mental model (30 sorts)
Small detected groups (2 + 2 sorts)
No detected mental model (2 sorts)
Action labels (7 of 36)
6
1
0
Frequency or necessity labels (5 of 36)
5
0
0
Device-control labels (5 of 36)
2
3
0
From grouping evidence to design decisions
Home-security and entry cards produced the strongest pairwise agreement, but no pair was grouped together in more than 28 of the 36 sorts. Mental Model Detection still found one dominant detected mental model across 30 sorts. A shared overall pattern does not mean every card has one obvious home.
Some cards contain several cues at once: a verb, a room and a topic. This study cannot isolate which cue influenced a particular grouping.
The next step is not to turn the most popular groupings directly into a menu. A product team can use the evidence to propose navigation options and then test whether people can find important functions. These three examples are hypotheses for tree testing, not navigation choices already validated by the card sort.
Three findings, three decisions to test
Research evidence
Candidate product decision
Tree-test check
28 of 36 sorts placed "Check whether a window is open" with "Check whether the garage door is closed."
Try a prominent entry/security status grouping that connects related checks.
"You're away and want to find out whether a window was left open."
"Manage your child's bedroom devices" had no partner above 15 of 36 sorts.
Compare access by room, household member and routine rather than assuming a single obvious home.
"You want to change the controls for devices in your child's bedroom."
Different label types appeared within the dominant detected mental model; T2 examines 16 of 36 sorts.
Test candidate category labels without assuming each label type requires a separate information structure.
Run equivalent tree-test tasks on the same structure with alternative category labels, then compare findability.
This case shows what respondents grouped together when they worked with everyday smart-home activities. A follow-up study will investigate a different question: whether activities with explicit household roles, timing and automation conditions bring out more distinct organizing principles.
Run your own open card sort
Let respondents create and name their own categories. UXbeam shows how they grouped the cards, while Mental Model Detection helps reveal similarities across completed sorts.
