Pavel Menkov
← All case studies

Shifted

UX/⁠UI case study · Android

AI-⁠powered used car marketplace

An online platform for buying and selling cars where every car and seller is automatically checked for honesty

Role
Researcher and designer
Platform
Android
Research
20 interviews, 3 competitors

01 · Goal and role

Is there a user segment that needs this kind of service

Business goal

Assess the market potential of a service for buying and selling used cars and determine whether there is a user segment whose need for such a service is strong enough to launch an MVP

My role

  • Designed and ran qualitative research: 20 in-⁠depth interviews with used car buyers to uncover their buying scenarios and key barriers
  • Analyzed 3 competitors (Drom, Auto.ru, Avito) to understand their value proposition, user scenarios and unmet audience needs
  • Built a CJM from the insights, identified problems and their solutions, formulated product hypotheses and proposed an MVP focused on simplifying seller and car verification
  • Designed the interface of the mobile app with the AI assistant

02 · Problems

Buying a used car is one big problem, and searching is the hardest part

  • 16 of 20

    respondents fear buying a low-quality car through online listings

  • 6 of 20

    wasted a lot of time driving to see a car in much worse condition than expected

  • 1.5 hours

    is the average drive to a car from a listing, only to be disappointed. And the same way back

  • 01

    No trust in listings

    People are afraid of being cheated: the seller hides the car’s flaws, sells it, and then it is unclear what to do

  • 02

    Lots of resellers

    They inflate the price for resale without improving the car’s condition and may hide its real history

  • 03

    Search takes a lot of time

    You have to drive to see every listing you like. Sometimes far away. And the car may turn out to be bad

  • 04

    Paid reports

    They cost around €2 each and still cannot guarantee the quality of the car

03 · Research

The user journey and two audience groups

CJM

The problems are clearly visible: satisfaction drops sharply during the car search

  1. Trigger

    Goal
    Improve driving comfort for myself and/⁠or my family
    Emotions
    Excitement about the upcoming purchase
    Pain points
    Anxiety over major financial choices & analysis paralysis
    Touchpoints
    Social media auto creators, personal recommendations from friends, and advertising

  2. Research & consideration

    Goal
    Explore market options
    Emotions
    Realizing that finding a car is going to be difficult
    Pain points
    Dishonest sellers, poor car condition, and wasted time
    Touchpoints
    Avito, Auto.ru, Drom, Facebook listings, dealers, knowledgeable friends

  3. Research & consideration

    Goal
    Buy a car safely without being scammed
    Emotions
    Complete frustration with the process
    Pain points
    Fear of finding nothing good and buying a lemon
    Touchpoints
    Avito, Auto.ru, Drom, Facebook listings, dealers, knowledgeable friends

  4. Connecting with the seller

    Goal
    Close the sale/⁠purchase deal
    Emotions
    Relief and joy after finding a good car
    Pain points
    Thought: “What if it’s a scam anyway?”
    Touchpoints
    In-⁠person communication

Personas

User interviews helped split the audience into 2 groups: young novice drivers open to new technologies, and adult experienced drivers who prefer familiar patterns

  • Alina

    Senior marketing manager

    Experience

    • Buying her first car
    • New driver

    Behavior

    • Uses natural language queries
    • Hasn’t decided on a car model yet

    Pain points

    • Afraid of getting scammed when buying a car

    How the product responds

    Search by text query and AI check of listings

  • Viktor

    Industrial mechanic

    Experience

    • Has had a lot of cars
    • Experienced driver

    Behavior

    • Searches for cars using specific filters
    • Knows exactly what car he needs

    Pain points

    • Fears wasting time: driving to see a car only to find it in bad shape

    How the product responds

    Search by parameters and a risk tag on the car card

04 · Competitors

Competitors do not solve the problem

Only Auto.ru comes close, but paid reports are still expensive

  • Avito

    Vehicle features

    • +Paid history reports available
    • −General marketplace

    Listings

    • −Dominated by unofficial resellers
    • −Poor quality of listings

    UX/⁠UI

    • +Simple and recognizable, but lacks focus on cars

  • Auto.ru

    Vehicle features

    • +Paid history reports available
    • +Dedicated automotive platform

    Listings

    • −Dominated by unofficial resellers
    • +High-quality listings

    UX/⁠UI

    • +Modern, well-crafted, and highly specialized

  • Drom

    Vehicle features

    • −Paid history reports are limited
    • +Dedicated automotive platform

    Listings

    • −Dominated by unofficial resellers
    • −Poor quality of listings

    UX/⁠UI

    • −Legacy design with overloaded interface

  • Shifted

    Vehicle features

    • +AI check of every listing
    • +Dedicated automotive platform

    Listings

    • +Sellers checked for suspicious activity
    • +Risks marked with a tag on the card

    UX/⁠UI

    • +Search by parameters or by text query

05 · Solution

A mobile app with the AI assistant

The insights led to a product solution: a mobile app with an AI assistant

  • 01

    Check

    Checks cars for hidden defects and sellers for suspicious activity

  • 02

    Search

    Helps find a car by turning a text query into filters. Write a request in free form and get listings

  • 03

    Inspection

    Helps check a car in person, replacing a professional car picker. Send photos to the chat and get an analysis of the car

06 · Search

I made search simple for both types of users. More conservative drivers can use standard filters, while tech-savvy newbies can search for cars by prompt

AI search

Alina

A newbie open to new technologies

  1. Home screen: a search bar, the ‘Car search options’ block with a switch between ‘By specifications’ and ‘AI Assistant’, and popular listings below

    01Home screen

  2. ‘Car search options’ in ‘AI Assistant’ mode: a request is being typed into the ‘Describe your dream car’ field, the keyboard is open

    02The request is typed in free form

  3. The request in plain words: budget, condition, color, body type and mileage. Below it, the ‘Search listings’ button

    03The request is entered, all that is left is to tap “Search listings”

Search by parameters

Viktor

An experienced driver used to filters

  1. ‘Search by specifications’, the ‘Make & model’ tab: a search field and a list of makes, BMW, Hyundai, Mercedes and others, with the ‘Apply’ button

    01Choosing make and model

  2. BMW models: a ‘Select all’ switch, series expand, and models are marked with checkboxes

    02Choosing models of one series, one by one or all at once

  3. ‘Search by specifications’, the ‘Price & year’ tab: sliders and ‘Min’ and ‘Max’ fields for the price and the year

    03Price and year range

07 · Results

Search results and refining the request

Search leads users to listings that match their parameters. The request can be extended however they like: through AI or through regular filters

Risk tag

AI immediately shows which car has risks

Risks are shown as a tag on the car card

  1. Search results: the request and the selected price and mileage filters at the top, car cards with the price, year and mileage below

    01Search results for the request

  2. Listing cards in the results: two cars have a red ‘Risks (3)’ tag

    02Risk tag on the car card

Refining with AI

Alina

The request can be changed with a message in the chat with the AI assistant

  1. Chat with the AI assistant: the user’s request, the reply about the listings found and the ‘Show 1,284 results’ button

    01Moving to the AI assistant from the results

  2. A new request about the price range is being typed in the chat with the AI assistant

    02A new request in free form

  3. The AI assistant updated the range to listings under €10,000 with any mileage and sorting turned off, and offers to show the results

    03The request is done and the range is updated

  4. New results with the ‘Min €700’ filter: Honda Vamos, VAZ Lada, VAZ Lada Granta and Daewoo Matiz

    04New results

Refining with filters

Viktor

Filters open as sheets over the results

  1. The ‘Price range’ sheet over the results: a slider, ‘Min’ and ‘Max’ fields and the ‘Show 1,284 results’ button

    01Price filter

  2. The ‘Mileage range’ sheet over the results: a slider and fields from 10,000 to 250,000 km

    02Mileage filter

  3. The ‘Sort by’ sheet: default, most popular, lowest price, highest price, newest listings and distance

    03Sorting

  4. The ‘Color’ sheet: an ‘Any’ switch and a list of colors with checkboxes

    04Color filter

08 · AI protector

The AI protector in listings

Listings have an AI check status. It shows that a car is safe or that it has risks

Safe listing

The “Verified listing” status

Photos match the description and the seller shows normal platform activity

  1. A Honda Vamos listing for €4,230 with the ‘Verified listing’ status: photos match the description and condition, the seller shows normal platform activity

    01Card with the “Verified listing” status

  2. Chat with the AI assistant about this listing: it explains that the photos match the description and the seller behaves normally

    02Breakdown in the chat with the AI assistant

Listing with risks

The AI finds mismatches

The red status shows the risks found, and the chat explains why they are dangerous

  1. The same listing with the red ‘3 risks found’ status: photos do not match the description, and the seller shows high sales activity

    01Card with the risks found

  2. The AI assistant explains the risks found, with the suggestions ‘What are the dangers?’ and ‘How to stay safe?’ below

    02The AI explains what does not add up

  3. The answer to ‘What are the dangers?’: the car may be worse than in the ad or have hidden problems. Solution: check the VIN and history, inspect the car in person and be careful with the seller

    03Why these risks are dangerous and what to do

09 · Details

Car details inside the listing

A listing shows similar cars, specifications with a 3D model, and features by category

Card sections

Everything about the car without leaving the listing

  1. Similar listings of the same model: four Honda Vamos cars with different prices and mileage, one with a ‘Risks (3)’ tag

    01Similar listings and cars

  2. The ‘Specifications’ tab: a 3D model of the car with its dimensions, then body type, engine, power, transmission, fuel type, drivetrain and fuel consumption

    02Specifications and a 3D model

  3. The ‘Features’ tab by category: safety, comfort, interior and exterior

    03Features by category

10 · AI assistant

The AI assistant in chat

The assistant lets you check a car in person on your own. Just take photos and the AI gives you a step-by-⁠step guide

Check by photos

From the first message to the full result

  1. AI assistant screen: the greeting ‘I’ll help you find your dream car’ and the suggestions ‘Check car by photo’ and ‘Match a car’

    01Standalone AI chat screen

  2. The ‘Check car by photo’ request: the assistant asks for photos, and the user sends two photos of the car

    02Car check by photos

  3. Preliminary assessment: no obvious issues with the exterior and interior, and photos of body gaps, bumpers, door sills and the dashboard are needed to finish the check. Confidence level 65%

    03Response and additional request

  4. Check updated after new photos: signs of front-end body repair found, the assistant recommends asking the seller about the damage history. Confidence level 87%

    04Full inspection results

11 · Stress test

Interface stress test

I separately checked the interface with extreme values: regular data on the left, the same screens with huge prices, mileages and counters on the right

Car card

Price, mileage, number of owners

  1. A listing card with regular values: the price €4,230, mileage 190,900 km and 3 previous owners

    01Regular values

  2. The same card with extreme values: the price €452,723,500, mileage 200,000,000 km and 999,999 previous owners fit without being cut off

    02Extreme values

Listing feed

Price, mileage, risk counter

  1. Results with regular values: prices from €1,500 to €58,900 and a ‘Risks (3)’ tag

    01Regular values

  2. Results with extreme values: prices €200,158,020 and €15,830,000, mileage 100,120,000 km and a ‘Risks (99,999)’ tag

    02Extreme values

Edge states

When nothing is found

  1. Empty results: ‘No listings found for these filters’ and buttons that expand the search

    01No listings for the selected filters

  2. Chat with the AI assistant: there are no listings under €10, and the assistant suggests raising the price to €700

    02No listings for the AI request

12 · Hypotheses and MVP

Five hypotheses and the MVP scope

The hypotheses are based on the research insights and tied to product features

  1. 01

    If the card shows the AI check status and a risk tag, buyers filter out doubtful listings before a trip and fear being cheated less

    No trust in listings

  2. 02

    If risks are visible before an inspection, the number of useless trips to bad cars goes down

    Search takes a lot of time

  3. 03

    If the check is built into the platform, buyers do not need outside reports for €2 that do not guarantee quality

    Paid reports

  4. 04

    If there is text query search next to the filters, novices find suitable cars faster, while experienced drivers keep their familiar path

    Two audience groups

  5. 05

    If the AI assistant analyzes photos of a car and gives a step-by-⁠step guide, buyers can inspect a car without a professional car picker

    Hidden defects

In the MVP

  • Search by filters and by text query with refining of results
  • A card with the AI check status and a risk tag
  • A listing with specifications and features
  • The AI assistant in chat: explaining risks and checking by photos

Not included, planned for later

  • Payment and the deal inside the platform
  • Inspection with a live expert
  • Extended tools for sellers

13 · Conclusion

A segment with a real need exists

The research showed that used car buyers have a strong need to verify listings: 16 of 20 respondents fear buying a low-quality car, and the existing competitor paths do not solve the problem. The project combines prompt-based car search with AI analysis of listings. The solution is meant to simplify search and make buying more transparent

  • 16 of 20

    respondents fear buying a low-quality car through a listing

  • 3

    competitors analyzed and none fully solves the trust problem

  • 5

    hypotheses the MVP is built on

More case studies