Lab · The Shortlist · 8 min read

The Shortlist

How might one shopping trip feel continuous across a phone, a store, and a person who works there? One trip, from a photo of a friend’s lamp to a store aisle, as a service map and a working prototype.

The problem
What a shopper works out at home stays on the phone. At the store door, the trip starts over.
What’s different
It carries the decision, not only the search: what you liked, your shortlist, and how sure the shelf count is.
What I built
A service map of one trip, and a working prototype at a phone’s width, on sample data.
What it shows
An idea worked through far enough to test. Nothing yet about demand, feasibility, or revenue: the study I’d run first is below.

An independent concept, designed and built by me with Claude Code, on sample data. No retailer is involved, the store is made up, and no shopper or employee has tried it yet.

Try it, it’s live

Shortlist

  1. At home
  2. In the store

What you found at home doesn’t follow you into the store

You like the lamp at a friend’s house, so you take a picture. That night you find four that look close and narrow it to two.

Two days later you’re in the store. Which two were yours? Your phone knows, the aisle doesn’t, and the person who works there starts with “What are you looking for?”

Every part of that trip works. What breaks is the space between them: the context doesn’t travel.

Searching by photo already ships. In September 2026 Target described Photo Search and Continue Shopping in its app. The question here comes after it: what should follow the shopper into the store?

One trip, from the shopper down to the systems

Switch to Connected to see where today’s trip breaks, and step through it one moment at a time.

Service map

What the shopper carries, and where it’s left behind

Illustrative: drawn from the scenario, not from observed shoppers. The systems are what a service like this would need, as I imagine them.

At home, on the phoneIn the store

  1. At a friend’s

    The shopper asks
    “Where would I find this?”
    How sure they feel
    Today: Fairly sureConnected: Fairly sure
    What they still have
    Photo
    TodayLine of visibility
    The photo sinks into a camera roll of four hundred.
    ConnectedLine of visibility
    The photo is kept as a question, with what you liked about it.
    Store employeeLine of internal interaction
    Not involved.
    Behind it
    Photo storage; visual search
    Where it can fail
    A poor photo leaves nothing to search.

  2. At home

    The shopper asks
    “Is this the one, or just close?”
    How sure they feel
    Today: UnsureConnected: Fairly sure
    What they still have
    Photo
    TodayLine of visibility
    Results look alike and don’t say why.
    ConnectedLine of visibility
    Each match says why, in words, and how it does on what you liked.
    Store employeeLine of internal interaction
    Not involved.
    Behind it
    Catalog attributes: shade, legs, height
    Where it can fail
    The catalog lacks the one attribute that matters.

  3. Saving two

    The shopper asks
    “Does my store have them?”
    How sure they feel
    Today: UnsureConnected: Fairly sure
    What they still have
    PhotoShortlistDoubts
    TodayLine of visibility
    “In stock,” with no date on it.
    ConnectedLine of visibility
    Each shelf count shows its age.
    Store employeeLine of internal interaction
    Counts the shelf on a schedule.
    Behind it
    Store inventory, with the time of each count
    Where it can fail
    There’s no timestamp to show.

  4. Walking in

    The shopper asks
    “Which two were mine?”
    How sure they feel
    Today: DoubtfulConnected: Sure
    What they still have
    PhotoShortlistDoubtsLeft at the door
    TodayLine of visibility
    The app opens on the storefront.
    ConnectedLine of visibility
    The app opens on the list in walking order, with each aisle and bay.
    Store employeeLine of internal interaction
    Not involved.
    Behind it
    The saved list; the planogram
    Where it can fail
    The planogram is stale and the bay is wrong.

  5. In the aisle

    The shopper asks
    “Settle, order, or leave?”
    How sure they feel
    Today: LostConnected: Fairly sure
    What they still have
    PhotoShortlistDoubts
    TodayLine of visibility
    One shelf is empty, and nothing says so.
    ConnectedLine of visibility
    A comparison, what you liked first, and three ways forward.
    Store employeeLine of internal interaction
    Not yet. Nobody is paged until asked.
    Behind it
    Nearby stock; order and delivery
    Where it can fail
    The alternatives feel like upselling.

  6. With a person

    The shopper asks
    “Do I explain it all again?”
    How sure they feel
    Today: LostConnected: Sure
    What they still have
    PhotoShortlistDoubts
    TodayLine of visibility
    They do. The employee starts from zero.
    ConnectedLine of visibility
    The employee sees only what the shopper chose to share.
    Store employeeLine of internal interaction
    Checks the back room; weighs in on the near match.
    Behind it
    A request routed to whoever in Home is free; the shopper’s consent
    Where it can fail
    Nobody is free. The request stays open, and the desk is named.

What the map shows

  • The trip breaks at the store door

    At every moment, a system knows something a person can’t get at. At the door, the shortlist stays behind on the phone.

    Moments 04 to 06

  • A timestamp does more than any screen

    “In stock” with no date is a guess. The time of each count tells the shopper how sure to be.

    Moments 03 and 05

  • The employee’s count makes the help possible

    Counting the shelf at moment three is what lets someone help at moment six, without starting from zero.

    Moments 03 and 06

Five decisions, each resting on a guess

The AI matches, explains, and suggests; the shopper decides the rest.

  1. Carry the context, not the cart

    The store gets the photo, what you liked about it, and your shortlist. Never your account or your history.

    My guess: what you liked is the one thing worth asking, once, under the photo.

  2. Say why, and how sure

    Each match says in words what matched and what didn’t: “Same shade, different legs,” not “87% match.”

    My guess: a sentence someone can check against the photo earns more trust than a score.

  3. Make “no” cheap

    “Not it” sits beside “Save,” the same size, and hiding all four is allowed: “none of these” is an answer.

    My guess: what someone turns down tells the app what they meant.

  4. Give every shelf count its age

    Three counted this morning is a fact. One last counted two days ago is closer to a guess, and the app says so before anyone drives four miles on it.

    My guess: the age, printed every time, changes whether someone makes the trip.

  5. Hand off to a person on purpose

    “Ask someone in Home” is on the store screen from the start, and you choose what the employee sees.

    My guess: an employee wants a stranger’s shortlist, and has the time for it.

Built for anyone in the aisle. Each new screen is announced and takes focus, nothing rests on color alone, and every control is at least 44 pixels. No screen reader user has tried it yet.

What I’d test first, and what would prove it wrong

The bet: if a shopper’s choices follow them into the store, they find things faster, decide with more confidence, and abandon fewer trips. For a retailer, that’s sales, satisfaction, and staff time.

The riskiest guess comes before all five decisions: does someone in an aisle want their phone out at all? If not, the in-store screens mostly go away. The app tells them which aisle to walk to and hands their shortlist to an employee.

With shoppers
Five people, each with their own photo, taking the whole trip with one of their two picks missing from the shelf.
With staff
Three people from a home department, fifteen minutes each, walked through the request.
With engineering and store operations
Which of the map’s bottom lanes exist today, and how old a shelf count really is.

  • Keep it

    If all three are true:

    • Most shoppers open the list in the aisle without being asked
    • Most can say why a match was suggested
    • A shelf count’s age changes at least one shopper’s mind

  • Simplify it

    If any one is true:

    • A plain list with aisle numbers works just as well
    • Shoppers ignore what they said they liked
    • Shoppers take the shelf count as a promise

  • Cut the in-store screens

    If either is true:

    • Shoppers only take their phone out to pay
    • Staff say the requests take more time than they save

An afternoon to build, twenty years to know what to build

The service map took a morning and the prototype an afternoon. What went into them took longer:

  1. The Pixel Farm studio: exposed brick, a hand-painted Pixel Farm sign on the wall, hanging lamps, and a pool table among the plants.

    2005 to 2013

    Eight years in studios

    The last five at Pixel Farm, where I built the sites I designed. That’s where I learned what a design costs to build.

  2. Under the line “Better experiences. Stronger enterprise.”, Adam at a wall screen showing an SAP home page of orders and approvals, walking a seated team through it.

    2013 to 2026

    Thirteen years at RBA

    Making 40+ SAP apps one product and a federal coalition’s grant tools one platform. I learned that people get lost between apps that each work on their own.

  3. A dusty slate crane hook hanging from a black cable, lifting a plain cream panel, and one taut black tag line running from the panel’s corner off the bottom of the frame, steering it.

    Now

    Coding agents

    They’re why the prototype took an afternoon. Agent Review is my experiment in the part left to me: deciding which of their options should ship.

What I’d still need to learn is the store itself: its shoppers, operations, merchandising, and economics.

What I’d bring is the part that didn’t get easier: finding the right problem, making it real enough to try, and helping a team decide what to build.