M.Des Thesis · IIT Delhi Industry Collaboration · Havells India · 2023

Cozy
Nights

A feature ask reframed as a systems problem. The brief was a sleep mode for an AC. The solution was a four-layer intelligence architecture inside the Havells Sync smart home app — saving roughly 18% on energy at equivalent comfort, while reducing night-time wake-ups.

Degree M.Des — Industrial Design
Institution IIT Delhi
Duration 5 months · 2023
Faculty Advisor Prof. Charu Monga
63
Survey Respondents
16
Pilot Interviews
2
Prototype Iterations
Context & Role

The short
version

Havells's brief was straightforward: design a sleep mode for the Havells Sync smart home app. I returned with a different problem. 62% of urban Indians wake up at least once during the night because the AC setpoint they chose before bed no longer matches their body temperature at 3 AM. 92% of them run a fan alongside the AC. 65% of them worry about what this is doing to their electricity bill. The feature ask was real, but it was the symptom of a deeper systems problem.

I designed a feature called Intellisleep, inside the Havells Sync app, that couples the AC and fan to maintain perceived thermal comfort through the night, learns from user behaviour over time, and closes its own feedback loop through a morning sleep-quality check-in. It lets people run their AC at 27 degrees with a fan and still feel like 24, saving roughly 18% on electricity while reducing night-time wake-ups.

The actual contribution is not the app feature. It is the systems reframe that produced it — and the diagnostic discipline of refusing to ship the brief as written when the brief was solving the wrong problem.

My role

I was the sole designer on this project, working end to end from research to final prototype — literature review across 80+ papers, 16 stakeholder interviews, a 63-person user survey, competitive analysis of 9 smart home energy products, concept development across four distinct directions, full system architecture, and two iterations of the Havells Sync app feature.

Collaboration

I worked within the Havells CXD team, reporting to Ananya Vetal (AGM, CXD) and mentored by Vineet Sharma (Asst. Manager, CXD). The Havells engineering team validated technical feasibility. Senior designers at Havells will carry the work forward into production.

Solution must be software-only, working with Havells' existing connected devices — integrated into the Havells Sync app architecture, respecting the Indian electricity-cost context.

The Problem

Why sleep breaks
at 3 AM

Most AC "sleep modes" today work like a simple timer. Every hour or two, the AC raises the setpoint by half a degree. This is a rough approximation of something real, but it misses the underlying physiology.

When you set your AC to 22 degrees at 11 PM, you have just eaten dinner. Your body is warm, digestion is active, and 22 degrees feels right. By 3 AM, digestion is complete, your metabolic rate has dropped, and your core body temperature has fallen as part of your circadian rhythm. The room is still at 22 degrees. Your body is not. The gap between them is why you wake up shivering, reach for a blanket, or stumble to the AC to change the setting.

Existing sleep modes cannot solve this because they don't model the body. They model the clock.

And there is a second problem stacked on top. Indian households are acutely cost-conscious about AC usage. The common workaround is that someone in the family wakes up at 2 AM to switch the AC off entirely. One person's sleep is sacrificed to protect the electricity bill. Every night. The same person, usually.

Two problems, one system: Devices operate on a fixed setpoint while bodies operate on a shifting circadian curve. And users cannot manually intervene during the night without sacrificing the very sleep the system is supposed to preserve.

Research

What the
data told us

Stakeholder interviews — 16 participants

The original project brief was about energy monitoring. Interviews with engineering, sales, and marketing leads at Havells surfaced something different: the most universally painful HVAC problem is not daytime comfort. It is sleep. That pivot shaped the rest of the project.

User survey — 63 respondents, Delhi NCR

62%
woke up at least once during the night
92%
used a fan and AC in combination
65%
reported cost anxiety around AC + fan use

Literature — three load-bearing citations

India Cooling Action Plan (ICAP, 2022) Only 8% of Indian households own an AC today. That number is expected to cross 40% within 20 years — making the thermal comfort problem a national-scale design opportunity.
Bureau of Energy Efficiency "AC @ 24" Campaign Every 1°C increase in AC setpoint saves approximately 6% on electricity. The system uses this to quantify the value of the coupled AC + fan approach.
TERI energy savings data An AC at 27°C with a ceiling fan produces perceived comfort equivalent to 24°C, because the fan creates evaporative cooling on the skin. Combined savings: roughly 18%. This was the insight that made the intervention possible — the AC and fan were already in the room. They just weren't talking to each other.
The Reframe & System

From feature
to system

I kept getting stuck trying to design a better AC remote. The shift happened when I stopped looking at the AC and the user in isolation, and started looking at the three subsystems that produce the sleep experience.

Subsystem 01
Physiological
Circadian rhythms, sleep stage architecture, REM sensitivity to thermal disruption. The body is doing what bodies do — on a predictable schedule.
Subsystem 02
Thermodynamic
The room as a thermal volume, the AC as a heat pump, the fan as an evaporative cooling aid. Each doing its job — but not coordinated.
Subsystem 03
Behavioural
Electricity-cost anxiety, household negotiation, Indian domestic sleep practices. Rational individual decisions producing irrational system-level outcomes.

No single subsystem is broken. The disruption emerges from the fact that they are not coupled. The design question stopped being "how do we design a better AC interface?" and became: "how do we introduce dynamic coupling across subsystems using hardware that is already in the home?"

Intellisleep — four layers

Layer 01
Sensing
The AC and fan discovered and paired over Bluetooth — using the BLE module already in these devices but sitting idle after first-time setup. The room becomes instrumented.
Layer 02
Actuation
The AC handles thermal load. The fan handles perceived temperature through airflow. Running in coupled mode, they deliver 24°C comfort at the energy cost of 27°C operation.
Layer 03
Adaptation
A rule-based algorithm at launch that learns the user's actual comfort profile over roughly a week. If you wake up at 3:15 AM three nights in a row and bump the temperature up, it starts pre-empting the adjustment on night four.
Layer 04
Feedback
The SATED morning check-in (Satisfaction, Alertness, Timing, Efficiency, Duration) — five questions, 30 seconds. The system's only window into whether the intervention worked, since the user was unconscious during the event being optimised.
The app is the handle. The system is the architecture.
The Interface

What the
user sees

The user never sees the four layers. They see a mode on their AC's device page, or a card on the Automations screen. The first time, they name a group ("Master bedroom"), the app scans for nearby AC and fan over Bluetooth, they set a preferred comfort temperature using a dial, and they choose the time window. Every night after, it runs on its own.

Key interaction design decisions

Bluetooth auto-discovery over manual device selection Most Havells smart devices have BLE modules that go unused after first-time pairing. Repurposing them for in-room discovery removed the painful step of scrolling through 15 devices to find the right two.
Dial-based temperature selection, not +/− Users often don't know their preferred sleep temperature in absolute terms. The dial made it feel like a feeling, not a number.
Two discovery paths, not one New users find it on the device page; power users find it under Automations. Both paths mapped onto the existing app structure so the feature doesn't feel bolted on.

V1 → V2 — what changed

V1 was too much like a toggle. Users didn't know what it was going to do. The mode card didn't read as interactive. V2 added a feedback state showing "AC and Fan linked" with a green tick, a micro-animation on the on/off transition so the state change was legible, and "how this works" hints for first-time users that disappeared after first use.

The "feels like" recommendation

One finding I took to the Havells engineering team (now in their backlog): Indian users don't think in degrees. They think in whether they feel hot or cold. My recommendation was to show both the absolute temperature and the "feels like" temperature — the way weather apps do. When the AC is at 27 and the fan is on, the app would tell the user "27° · feels like 24°." This bridges the thermodynamic reality and the user's mental model, and opens a conversation about cost savings without lecturing.

Scope note: Indian bedrooms are shared. One thermal field cannot optimise for a thin person and a heavy one, a child and a grandparent. Intellisleep reduces wake-ups for a representative sleeper — it doesn't solve the fundamental actuator mismatch in a shared room. That problem needs per-person actuation. The users still self-regulate at the last mile, through blanket thickness and one leg out.

Outcomes & Reflection

What was built,
what was learned

Handed off to Havells

Complete interactive prototype — Intellisleep in Havells Sync Full Figma prototype across both discovery paths (device page + Automations), group creation flow, Bluetooth scan state, temperature dial, and morning check-in.
System architecture across four layers Sensing → Actuation → Adaptation → Feedback, with the "feels like" recommendation and SATED check-in framework documented for the engineering team.
Future scope roadmap Per-person actuation pathways and the broader ecosystem — smart lights, curtains, air purifier — that can plug into the same Intellisleep automation architecture.
M.Des CGPA 8.0 — IIT Delhi Completed the M.Des programme with a CGPA of 8.0, combining thesis research with coursework in Applied Ergonomics, Design for User Experience, and Product Interface Design.

Three lessons

Lesson 01
Specific contexts beat generic users
Every finding in this project became sharper when I stopped saying "the user" and started saying "an urban Indian adult sleeping in a shared bedroom with a split AC and a ceiling fan." The design got better in proportion to how specifically I could name the user.
Lesson 02
The hardest part of systems design is drawing the boundary
Everything is connected to everything. The hard discipline is saying: this is in scope, this is not, here is the coupling I am intervening at, here is what I am accepting as given.
Lesson 03
Feedback loops need unconscious observability
People cannot self-report on experiences they slept through. Morning-after frameworks like SATED are not a workaround. They are the right instrument for the problem.
What this proves
Diagnostic discipline before execution
The most consulting-coded move on this project was refusing the brief as stated. The client asked for a sleep mode. The diagnostic work showed the problem wasn't a feature gap — it was a coupling failure between two devices and a body. Reframing the brief is the highest-leverage thing a strategist does. This was the first time I did it formally, and it shaped how I work inside enterprise programmes today.

Credits

Faculty Supervisor
Prof. Charu Monga
Department of Design, IIT Delhi
Industry Supervisor
Ananya Vetal
AGM, CXD, Havells India
Industry Mentor
Vineet Sharma
Asst. Manager, CXD, Havells India
Special Thanks
Ekta Shrivastava
And the wider Havells UX team