Proptech
SCAN TO SCOPE
A self-initiated concept for Zero Homes, a Denver home electrification company. I designed one shared record of the home that carries a homeowner's phone scan through engineering review to install day: a home profile that coaches the homeowner before the camera opens, an incentive meter that rewards every finished step with real money, a guided scan that checks its own coverage before upload, a design review where an AI agent drafts and the engineer decides, and a contractor app that checks only what a camera can't see. As the solo product designer, I recorded and studied the live app, read Zero's customer terms, and built a clickable prototype of all three apps.
My Role
My Role
Product Designer (Solo, Self-initiated Concept)
Industry
Industry
Home Development
Contributions
Contributions
UX Research / AI Agent Design / Mobile & Web UX/UI / Service Design / Prototyping
Scan to Scope: Making a Remote Home Scan Trustworthy Enough to Survive Install Day
Scan to Scope: Making a Remote Home Scan Trustworthy Enough to Survive Install Day
Overview
Overview
CHALLENGE
Zero Homes replaces the in-home sales visit with a phone scan. Engineers turn that scan into a heat pump design and quote (Zero Homes), and the crew's plan and install day inherit whatever the scan caught or missed. Under Zero's own customer terms, extra work found on install day is Zero's cost; only concealed conditions get a price change, agreed in writing. Volume is climbing fast: revenue grew 750% in 2025 (ColoradoBiz). The job posting names three products to design: the homeowner scan app, a new contractor app and a web HVAC design tool (Zero Homes).
The design question: how do we make a remote capture trustworthy enough that the quote survives the install?
Zero Homes replaces the in-home sales visit with a phone scan. Engineers turn that scan into a heat pump design and quote (Zero Homes), and the crew's plan and install day inherit whatever the scan caught or missed. Under Zero's own customer terms, extra work found on install day is Zero's cost; only concealed conditions get a price change, agreed in writing. Volume is climbing fast: revenue grew 750% in 2025 (ColoradoBiz). The job posting names three products to design: the homeowner scan app, a new contractor app and a web HVAC design tool (Zero Homes).
The design question: how do we make a remote capture trustworthy enough that the quote survives the install?
SOLUTION
I designed one shared record of the home that moves from the homeowner's phone to the engineer's screen to the crew's truck. Each user sees it in the form they need:
Homeowner app: an incentive meter that fills with real money as each step finishes, a short home profile that coaches the homeowner and sets verified truths before the camera opens, a scan plan built from the answers (13 rooms on 3 levels, ~17 min), a scanner that coaches instead of scolds, a scan check that flags weak areas before upload, and equipment photos the agent reads for you.
Engineer design review: the agent drafts the room-by-room load calculation and matches incentives, with a source and confidence on every input. Low confidence blocks approval until an engineer decides.
Contractor app: offers that show verified scope and risks up front, two on-site checks instead of a full re-survey, change orders tagged by who pays under Zero's terms, and rebate photos as a checklist before the crew leaves.
The design goal: the agent drafts, people decide. Every number shows where it came from, and anything a camera can't see goes to a person, not a guess.
I designed one shared record of the home that moves from the homeowner's phone to the engineer's screen to the crew's truck. Each user sees it in the form they need:
Homeowner app: an incentive meter that fills with real money as each step finishes, a short home profile that coaches the homeowner and sets verified truths before the camera opens, a scan plan built from the answers (13 rooms on 3 levels, ~17 min), a scanner that coaches instead of scolds, a scan check that flags weak areas before upload, and equipment photos the agent reads for you.
Engineer design review: the agent drafts the room-by-room load calculation and matches incentives, with a source and confidence on every input. Low confidence blocks approval until an engineer decides.
Contractor app: offers that show verified scope and risks up front, two on-site checks instead of a full re-survey, change orders tagged by who pays under Zero's terms, and rebate photos as a checklist before the crew leaves.
The design goal: the agent drafts, people decide. Every number shows where it came from, and anything a camera can't see goes to a person, not a guess.
RESEARCH
RESEARCH
I set up Zero's live app (version 3, released October 2026) and scanned part of my own home in Denver. It already does a lot right: tips you must read before you can start, camera and location explained before the system asks, and a "too close" warning that uses text and pattern, not color alone. The gaps showed up after the camera opened:
A 22-second scan of one room was marked "Ready to upload." Nothing told me the kitchen, basement and garage were missing.
The tips are the same for every home. No room list, no hint where the furnace or panel lives, no time estimate for this house. The homeowner is left on a scavenger hunt.
Live feedback says what's wrong, not how to fix it. Getting close to a wall fills the screen with "TOO CLOSE" stripes.
Upload is a separate manual step. The phone has to stay unlocked in the app for 3 to 8 minutes, so a finished scan can sit on the phone and never reach an engineer.
The primary button starts over. After closing the upload sheet, the big button says "Start a home scan," while Upload is a small pill.
I also read Zero's customer terms. Section 6.1 makes install-day surprises Zero's cost, which turned "is the scan complete?" from a UX question into a business one. This was one session on one home, and Zero may check scans after upload in ways I couldn't see.
Where AI helps, and where it falls down
AI agents are great at reading and math from clean inputs, and bad at anything a camera can't see.
Agent: reads data plates and panel labels, measures rooms and windows from the scan, drafts the load calc, matches incentives. Every value shows the photo or scan it came from.
Engineer: insulation inside walls, the panel plan, system size and price. Low-confidence inputs block approval until an engineer accepts or edits them.
Contractor on site: return duct size and line-set route, hidden from the camera. A guess here becomes a change order.
How I'd measure it
Targets I'd set before shipping, not results. The concept hasn't been tested with users yet.
Scans complete on first upload: up
Scans finished but never uploaded: down
Engineer callbacks or rescan requests per scan: down
Install-day change orders, split by Zero-paid and concealed: down
Return trips for missing rebate photos: down
How often engineers override the agent's high-confidence inputs: if it's high, the labels aren't earning trust yet
How I built it
I used AI to build faster, not to decide. I recorded the live app and read Zero's terms myself, and used Claude to break down the recording and check my claims against sources. I built the prototype in Perplexity, screen by screen. The framing, the flow and every tradeoff are my calls. Try all three apps here:
I set up Zero's live app (version 3, released October 2026) and scanned part of my own home in Denver. It already does a lot right: tips you must read before you can start, camera and location explained before the system asks, and a "too close" warning that uses text and pattern, not color alone. The gaps showed up after the camera opened:
A 22-second scan of one room was marked "Ready to upload." Nothing told me the kitchen, basement and garage were missing.
The tips are the same for every home. No room list, no hint where the furnace or panel lives, no time estimate for this house. The homeowner is left on a scavenger hunt.
Live feedback says what's wrong, not how to fix it. Getting close to a wall fills the screen with "TOO CLOSE" stripes.
Upload is a separate manual step. The phone has to stay unlocked in the app for 3 to 8 minutes, so a finished scan can sit on the phone and never reach an engineer.
The primary button starts over. After closing the upload sheet, the big button says "Start a home scan," while Upload is a small pill.
I also read Zero's customer terms. Section 6.1 makes install-day surprises Zero's cost, which turned "is the scan complete?" from a UX question into a business one. This was one session on one home, and Zero may check scans after upload in ways I couldn't see.
Where AI helps, and where it falls down
AI agents are great at reading and math from clean inputs, and bad at anything a camera can't see.
Agent: reads data plates and panel labels, measures rooms and windows from the scan, drafts the load calc, matches incentives. Every value shows the photo or scan it came from.
Engineer: insulation inside walls, the panel plan, system size and price. Low-confidence inputs block approval until an engineer accepts or edits them.
Contractor on site: return duct size and line-set route, hidden from the camera. A guess here becomes a change order.
How I'd measure it
Targets I'd set before shipping, not results. The concept hasn't been tested with users yet.
Scans complete on first upload: up
Scans finished but never uploaded: down
Engineer callbacks or rescan requests per scan: down
Install-day change orders, split by Zero-paid and concealed: down
Return trips for missing rebate photos: down
How often engineers override the agent's high-confidence inputs: if it's high, the labels aren't earning trust yet
How I built it
I used AI to build faster, not to decide. I recorded the live app and read Zero's terms myself, and used Claude to break down the recording and check my claims against sources. I built the prototype in Perplexity, screen by screen. The framing, the flow and every tradeoff are my calls. Try all three apps here:
PROCESS
PROCESS
Micro-coaching: verified truths before the camera opens
Not every house is the same, but today every homeowner gets the same five tips and then a camera and a timer. Where is the panel? Does the garage count? Is the basement done? The homeowner is on a scavenger hunt with no map, and nothing tells them when they've found everything.
I'd solved this before. Before ReaLift scans a foot, it asks three questions: the shopper's usual shoe size, their sizing system (US, EU, UK or JP) and their sex. Those answers are verified truths the scan is checked against, and the questions coach the shopper before the hard part starts. Our backend engineers asked for the most specific input possible, because it left no room for error.
Scan to Scope does the same for a house:
Confirm, don't retype. Public-record facts (year built, square feet, bedrooms, bathrooms) are pre-filled. The homeowner taps "Yes, that's right" or "We finished the basement."
Ask only what records can't know: levels, basement, heating type, where the panel is.
Coach while asking. Each question says why it matters ("Confirms the state heat pump credit"). The panel question sets expectations about 100 A panels before the quote does.
Turn answers into a map: 13 rooms on 3 levels in walking order, where the furnace and panel usually live, and "~17 min for your home."
The verified truths also power the scan check. The app knows this house has 13 rooms, so it can say the stairs are 40% captured instead of accepting a 22-second scan. It's the same job ReaLift's credit card did: one trusted reference that everything else is checked against.
Gamification with real stakes: the incentive meter
The heart of the homeowner app is a meter that fills with real money. Each finished step confirms an incentive the home qualifies for, from $1,500 at 2 of 5 steps to $6,450 at the end (sample amounts). Every step says what it unlocks before you start it, and every stamp on the meter opens its step. People are more willing to keep going when they know they're being rewarded and can see the reward getting closer.
Expectancy theory (Vroom, 1964). From my I/O psychology training: people put in effort when they believe effort leads to a result, the result leads to a reward, and they value the reward. The meter makes all three visible: a ~2-minute step, a named credit, money for their own home.
Endowed progress. Account and address already count, so you start at 2 of 5. Car-wash cards with 2 of 10 stamps pre-filled were finished by 34% of customers, versus 19% for blank 8-stamp cards (Nunes & Drèze, 2006).
Goal gradient. Café customers bought coffee faster as a free one got closer (Kivetz, Urminsky & Zheng, 2006). The meter and one big next-step card keep the finish line in view.
Open loops. A short, visible, unfinished checklist keeps pulling people back, as long as "done" is obvious (Zeigarnik effect).
Guardrails. A meta-analysis of 40 experiments found gamification has small positive effects, and the motivational effect didn't hold in the most rigorous studies (Sailer & Homner, 2020), so this uses real rewards, not badges. Expected rewards can undermine motivation for tasks people already enjoy (Deci, Koestner & Ryan, 1999, 128 studies); scanning a basement isn't one of them. Every stamp is a real step, and an amount only appears once eligibility is confirmed.
Screen by screen
I kept Zero's tips tutorial as it is and redesigned the flow around it.
A checklist that pays you back
Today: Home Set Up is a sideways carousel of 4 cards; steps 3 and 4 say "Create account to unlock," and step 4 is a 30-minute advisor video call.
What changes: One vertical list under the incentive meter, each step showing its time and what it unlocks. One big next-step card, a centered confirmation after every step, and the advisor call moves to the end as an option.
Why: The theory above. Progress you can see, steps you can reopen, and the next move is never hidden.
Verified truths before the camera opens
Today: After the address, the app goes straight to five generic tips. Nothing asks about levels, basement, heating type or the panel.
What changes: A 6-question home profile with public-record facts to confirm, each question explaining why it matters.
Why: Every answer cuts guesswork for the engineer doing the load calc, and drives the route, equipment list and time estimate.
Your scan plan, built from your answers
Today: No list of rooms, no hint where equipment lives, and a generic "~10 min."
What changes: Rooms in walking order by level, equipment with where-to-look hints, three prep tasks, and a time estimate for this house.
Why: The same list becomes the live tracker during the scan, so the homeowner always knows what's left.
A scanner that coaches instead of scolds
Today: A raw camera feed and a timer. Getting close to a door fills the screen with red "TOO CLOSE" stripes, and stopping just asks "Are you sure?"
What changes: A room label, a coverage ring and a mini floor plan that fills in as you walk. Say "I'm in the kitchen" and the room switches. "Too close" becomes "step back about 2 feet," and the scan pauses instead of failing.
Why: Feedback in the moment is cheaper than a callback later, and an error message should say how to recover, without alarm.
Catch gaps while they're still home
Today: The upload sheet says don't lock your phone or switch apps for 3 to 8 minutes. After closing it, the big button says "Start a home scan," while Upload is a small pill.
What changes: A scan check flags weak areas with a 30-second fix, and "I can't get to this area" is a valid answer passed to the engineer. Upload runs in the background, with one primary action: continue.
Why: Fixing the stairs now takes 30 seconds. Fixing it after upload takes a call, a reschedule or a guess.
Equipment photos that read themselves
Today: A separate ~15-minute photo walk-through, locked until the account exists.
What changes: Only the equipment this home has, required items first, with a framing guide each. The agent reads the data plate and shows what it found with a plain-language confidence label. Glare gets a specific retake tip.
Why: Panel capacity is the classic install-day surprise. Telling the homeowner "100 A, options in your quote" builds trust before the price.
A clear finish line
Today: Step 4 is a 30-minute advisor video call.
What changes: A done state showing what was unlocked, what happens next and a 24-hour quote window. Talk to an advisor now or book 15 minutes, if they want to.
Why: A clear done state closes the loop and is the reward people came for.
Mistakes you can fix
Today: No redo path short of starting over.
What changes: Every room has Re-record, which replaces only that room. A missed room can be added and scanned on the spot, and photos can be retaken.
Why: People finish more when a mistake is cheap to undo.
Micro-coaching: verified truths before the camera opens
Not every house is the same, but today every homeowner gets the same five tips and then a camera and a timer. Where is the panel? Does the garage count? Is the basement done? The homeowner is on a scavenger hunt with no map, and nothing tells them when they've found everything.
I'd solved this before. Before ReaLift scans a foot, it asks three questions: the shopper's usual shoe size, their sizing system (US, EU, UK or JP) and their sex. Those answers are verified truths the scan is checked against, and the questions coach the shopper before the hard part starts. Our backend engineers asked for the most specific input possible, because it left no room for error.
Scan to Scope does the same for a house:
Confirm, don't retype. Public-record facts (year built, square feet, bedrooms, bathrooms) are pre-filled. The homeowner taps "Yes, that's right" or "We finished the basement."
Ask only what records can't know: levels, basement, heating type, where the panel is.
Coach while asking. Each question says why it matters ("Confirms the state heat pump credit"). The panel question sets expectations about 100 A panels before the quote does.
Turn answers into a map: 13 rooms on 3 levels in walking order, where the furnace and panel usually live, and "~17 min for your home."
The verified truths also power the scan check. The app knows this house has 13 rooms, so it can say the stairs are 40% captured instead of accepting a 22-second scan. It's the same job ReaLift's credit card did: one trusted reference that everything else is checked against.
Gamification with real stakes: the incentive meter
The heart of the homeowner app is a meter that fills with real money. Each finished step confirms an incentive the home qualifies for, from $1,500 at 2 of 5 steps to $6,450 at the end (sample amounts). Every step says what it unlocks before you start it, and every stamp on the meter opens its step. People are more willing to keep going when they know they're being rewarded and can see the reward getting closer.
Expectancy theory (Vroom, 1964). From my I/O psychology training: people put in effort when they believe effort leads to a result, the result leads to a reward, and they value the reward. The meter makes all three visible: a ~2-minute step, a named credit, money for their own home.
Endowed progress. Account and address already count, so you start at 2 of 5. Car-wash cards with 2 of 10 stamps pre-filled were finished by 34% of customers, versus 19% for blank 8-stamp cards (Nunes & Drèze, 2006).
Goal gradient. Café customers bought coffee faster as a free one got closer (Kivetz, Urminsky & Zheng, 2006). The meter and one big next-step card keep the finish line in view.
Open loops. A short, visible, unfinished checklist keeps pulling people back, as long as "done" is obvious (Zeigarnik effect).
Guardrails. A meta-analysis of 40 experiments found gamification has small positive effects, and the motivational effect didn't hold in the most rigorous studies (Sailer & Homner, 2020), so this uses real rewards, not badges. Expected rewards can undermine motivation for tasks people already enjoy (Deci, Koestner & Ryan, 1999, 128 studies); scanning a basement isn't one of them. Every stamp is a real step, and an amount only appears once eligibility is confirmed.
Screen by screen
I kept Zero's tips tutorial as it is and redesigned the flow around it.
A checklist that pays you back
Today: Home Set Up is a sideways carousel of 4 cards; steps 3 and 4 say "Create account to unlock," and step 4 is a 30-minute advisor video call.
What changes: One vertical list under the incentive meter, each step showing its time and what it unlocks. One big next-step card, a centered confirmation after every step, and the advisor call moves to the end as an option.
Why: The theory above. Progress you can see, steps you can reopen, and the next move is never hidden.
Verified truths before the camera opens
Today: After the address, the app goes straight to five generic tips. Nothing asks about levels, basement, heating type or the panel.
What changes: A 6-question home profile with public-record facts to confirm, each question explaining why it matters.
Why: Every answer cuts guesswork for the engineer doing the load calc, and drives the route, equipment list and time estimate.
Your scan plan, built from your answers
Today: No list of rooms, no hint where equipment lives, and a generic "~10 min."
What changes: Rooms in walking order by level, equipment with where-to-look hints, three prep tasks, and a time estimate for this house.
Why: The same list becomes the live tracker during the scan, so the homeowner always knows what's left.
A scanner that coaches instead of scolds
Today: A raw camera feed and a timer. Getting close to a door fills the screen with red "TOO CLOSE" stripes, and stopping just asks "Are you sure?"
What changes: A room label, a coverage ring and a mini floor plan that fills in as you walk. Say "I'm in the kitchen" and the room switches. "Too close" becomes "step back about 2 feet," and the scan pauses instead of failing.
Why: Feedback in the moment is cheaper than a callback later, and an error message should say how to recover, without alarm.
Catch gaps while they're still home
Today: The upload sheet says don't lock your phone or switch apps for 3 to 8 minutes. After closing it, the big button says "Start a home scan," while Upload is a small pill.
What changes: A scan check flags weak areas with a 30-second fix, and "I can't get to this area" is a valid answer passed to the engineer. Upload runs in the background, with one primary action: continue.
Why: Fixing the stairs now takes 30 seconds. Fixing it after upload takes a call, a reschedule or a guess.
Equipment photos that read themselves
Today: A separate ~15-minute photo walk-through, locked until the account exists.
What changes: Only the equipment this home has, required items first, with a framing guide each. The agent reads the data plate and shows what it found with a plain-language confidence label. Glare gets a specific retake tip.
Why: Panel capacity is the classic install-day surprise. Telling the homeowner "100 A, options in your quote" builds trust before the price.
A clear finish line
Today: Step 4 is a 30-minute advisor video call.
What changes: A done state showing what was unlocked, what happens next and a 24-hour quote window. Talk to an advisor now or book 15 minutes, if they want to.
Why: A clear done state closes the loop and is the reward people came for.
Mistakes you can fix
Today: No redo path short of starting over.
What changes: Every room has Re-record, which replaces only that room. A missed room can be added and scanned on the spot, and photos can be retaken.
Why: People finish more when a mistake is cheap to undo.
DECISIONS
DECISIONS
Design review: the agent drafts, the engineer decides
When the scan arrives, the engineer opens one web screen: the floor plan with heat loss per room, the equipment the agent read from photos, and a first-pass load calculation. Every input carries its source and a confidence label.
High confidence is auto-accepted: window area from the scan, ceiling heights, the furnace read from its data plate.
Low confidence blocks approval: wall insulation assumed from the year built can't be seen by a camera, so the engineer accepts or edits it.
Hidden items go to the contractor, not to a guess: return duct size, the line-set route.
Real tradeoffs stay with the engineer: keep the 100 A panel with a load-management device (sample +$650), or upgrade to 200 A (+$3,800 and 2 to 4 weeks of utility scheduling).
"Approve design and send quote" stays disabled until every open decision is settled. It's my ReaLift HOLD button again: the action unlocks only when the inputs are ready.
Contractor app: two checks, not a re-survey
Offers that show the risk up front: "6 of 8 verified," with risks like a 100 A panel as chips. Contractors price risk, so verified scope means faster yeses and fewer padded bids.
The scan, translated for the crew: scope, site photos and on-site checks, with the agent's readout on each photo.
Two on-site checks instead of a 40-item survey, each answered yes or no.
Change orders without stopping the job: a voice note, a photo and a suggested price and time impact, tagged by who pays. Extra work found on install day is Zero's cost under its terms (section 6.1), so the homeowner's price doesn't change. Only a hidden condition, like damage inside a wall (6.1.1), can change the price, and only after the homeowner agrees in writing. A homeowner who sees "Paid by Zero" keeps trusting the quote.
Rebate paperwork as a checklist, done before the crew leaves, because a missing photo means a return trip.
An ending worth remembering: payment, rebates filed and zero return trips. People judge an experience mostly by its peak and its end (Alaybek et al., 2022).
Design review: the agent drafts, the engineer decides
When the scan arrives, the engineer opens one web screen: the floor plan with heat loss per room, the equipment the agent read from photos, and a first-pass load calculation. Every input carries its source and a confidence label.
High confidence is auto-accepted: window area from the scan, ceiling heights, the furnace read from its data plate.
Low confidence blocks approval: wall insulation assumed from the year built can't be seen by a camera, so the engineer accepts or edits it.
Hidden items go to the contractor, not to a guess: return duct size, the line-set route.
Real tradeoffs stay with the engineer: keep the 100 A panel with a load-management device (sample +$650), or upgrade to 200 A (+$3,800 and 2 to 4 weeks of utility scheduling).
"Approve design and send quote" stays disabled until every open decision is settled. It's my ReaLift HOLD button again: the action unlocks only when the inputs are ready.
Contractor app: two checks, not a re-survey
Offers that show the risk up front: "6 of 8 verified," with risks like a 100 A panel as chips. Contractors price risk, so verified scope means faster yeses and fewer padded bids.
The scan, translated for the crew: scope, site photos and on-site checks, with the agent's readout on each photo.
Two on-site checks instead of a 40-item survey, each answered yes or no.
Change orders without stopping the job: a voice note, a photo and a suggested price and time impact, tagged by who pays. Extra work found on install day is Zero's cost under its terms (section 6.1), so the homeowner's price doesn't change. Only a hidden condition, like damage inside a wall (6.1.1), can change the price, and only after the homeowner agrees in writing. A homeowner who sees "Paid by Zero" keeps trusting the quote.
Rebate paperwork as a checklist, done before the crew leaves, because a missing photo means a return trip.
An ending worth remembering: payment, rebates filed and zero return trips. People judge an experience mostly by its peak and its end (Alaybek et al., 2022).
NEXT STEPS
NEXT STEPS
What I'd do next
Interview Zero's engineers: how do they turn a scan into a design today, and where do they lose time?
Ride along on two installs to see which on-site surprises actually happen.
Test the scan check with five homeowners: do they fix a flagged room, or skip it?
Validate the contractor offer screen with partner contractors before building it out.
Unaffiliated concept. Not affiliated with Zero Homes. All people, addresses and dollar amounts in the prototype are fictional.
What I'd do next
Interview Zero's engineers: how do they turn a scan into a design today, and where do they lose time?
Ride along on two installs to see which on-site surprises actually happen.
Test the scan check with five homeowners: do they fix a flagged room, or skip it?
Validate the contractor offer screen with partner contractors before building it out.
Unaffiliated concept. Not affiliated with Zero Homes. All people, addresses and dollar amounts in the prototype are fictional.
Design
Design
IMPACT
IMPACT
For Homeowners & Contractors
Gaps get fixed in 30 seconds while the homeowner is still home, not with a callback or a reschedule. Crews arrive with two checks instead of a full re-survey, and rebate photos are done before they leave.
For Zero Homes
