EXACT MATCH · CONFIRM THE VEHICLE AND ACCELERATE ACTION
The match is explicit, trust evidence is elevated, and the primary action moves from generic contact to starting the purchase.
CARS.COM MARKETPLACE · STRATEGIC PRODUCT SYSTEM
A vehicle page that adapts tohow someone decides.
I used AI to explore shopper needs, then narrowed the work to two decisions: confirm this exact vehicle or find the best fit for a monthly budget. I designed one VDP whose order and next action could change without redrawing the whole page.
AI-ASSISTED DISCOVERY → ONE TRUSTED SHELL → TWO DECISION PATHS
AI-ASSISTED DISCOVERY
Two decision jobs, not two audience skins.
01SignalsSearch · views · saves
02AI synthesisPatterns · objections
03Decision jobsExact match · payment first
04Governed modesOrder · proof · action
PAYMENT FIRST · ORGANIZE AROUND THE CONSTRAINT
Monthly budget leads; tradeoffs and alternatives stay visible before a quote request.
SYSTEM
1 shellTRUSTED MARKETPLACE FOUNDATION
4 modesDEFAULT · CONTACT · EXACT · PAYMENT
2 jobsCONFIRM THIS CAR · FIND WHAT FITS
ReusableTEST SEAMS, NOT ONE-OFF PAGES
FROM ONE FIXED PAGE TO A GOVERNED DECISION SYSTEM
The work went beyond a better VDP.It turned one fixed page into a system.
Every shopper saw the same hierarchy, form, and module order. I separated the facts that had to stay stable from the decision support teams could configure and test.
BEFORE · INHERITED EXPERIENCE · NOT MY DESIGN
AFTER · GOVERNED DEFAULT
AI-ASSISTED DISCOVERY · HUMAN JUDGMENT
AI helped me find the useful difference.The personas became decision jobs, not profiles.
AI helped me compare behavior, surface likely objections, and pressure-test the audience ideas. I used the strongest patterns to define two decision jobs the team could recognize and test.
EXACT-MATCH SHOPPER
“I know the car. Help me confirm this one.”
- SIGNALS
- Precise make/model intent · repeat VDP views · same-model comparison
- LIKELY QUESTION
- Is this VIN available, trustworthy, and worth moving on now?
- PRODUCT RESPONSE
- Show why the vehicle matches, surface its history, and keep similar inventory close.
Precise query
Repeat views→Confirm this VIN
with confidence→History first
Start purchase
Repeat views→Confirm this VIN
with confidence→History first
Start purchase
PAYMENT-FIRST EXPLORER
“I know the number. Help me see what fits.”
- SIGNALS
- Monthly budget · price sensitivity · broader model set · under-budget alternatives
- LIKELY QUESTION
- Can I make this work without wasting time on a vehicle that does not fit?
- PRODUCT RESPONSE
- Lead with payment controls, make tradeoffs visible, and preserve the same budget in fallback inventory.
Monthly budget
Price sensitivity→Find what fits
my number→Payment first
Quote CTA
Price sensitivity→Find what fits
my number→Payment first
Quote CTA
ONE SYSTEM · TWO DECISION PATHS
The facts stay fixed.The sequence changes around the decision.
Both versions use the same vehicle, price, dealer, availability, and history. Only the order of proof, primary action, and fallback change to match the shopper’s decision.
TESTING ARCHITECTURE · CONFIGURE, DO NOT REDRAW
Each test changed one thing on purpose.That made the next experiment faster to build.
The shell stayed fixed. Each test changed only the modules tied to the hypothesis, which made versions easier to compare and the learning easier to reuse.
Audience hypothesis
What pattern might matter?
Decision job
What is the shopper resolving?
Governed mode
Which approved behavior fits?
Slot configuration
What order, proof, and action change?
Experiment + learning
What did the controlled change teach us?
THREE TESTS · ONE SHARED SYSTEM
Move history and same-model fallback ahead of financing.
Start my purchase
Move payment controls and under-budget inventory ahead of deep proof.
Request a payment quote
Keep the full lead form when seller contact is already the job.
Check availability
WHAT THE WORK REQUIRED
The hard part was decidingwhat personalization was allowed to change.
Default first
Strengthen the universal VDP before introducing model-driven states.
Modes, not themes
Change the page for the decision someone is making, not a marketing persona.
Explain the change
Make the recommendation, match, or budget logic visible to the shopper.
Protect trust
Preserve inventory truth, alternatives, reversibility, and dealer accountability.
Personalization earns its place only when it reduces decision work without reducing the shopper’s agency.
THE PROPOSAL
A marketplace decision system,not a collection of personalized pages.
The result was one dependable vehicle page with four approved modes. Teams could change hierarchy, proof, tools, alternatives, and calls to action without rebuilding the page or changing the underlying facts.
1 shellMARKETPLACE FOUNDATION
4 modesDEFAULT · CONTACT · EXACT · PAYMENT
2 jobsEXACT MATCH + PAYMENT FIRST
Test seamsORDER · PROOF · CTA · FALLBACK