Shift promised a stress-free, fully online way to buy a car. Customers told us it felt exactly like the dealership they were trying to avoid.

We didn’t guess at what was broken. The evidence was already sitting in four places nobody had put side by side: site analytics and heat maps, user interviews and surveys, CX emails, call logs, and chatbot transcripts, and the sales floor itself. That became the operating loop for every release.

1. Diagnose from the record — heat maps and drop-off analytics on the old pages, paired with what customers were literally typing into the chatbot and saying on support calls.
2. Rebuild the flow architecture first — map the starting state (online vs. offline) against the ending state (all online) before touching a screen.
3. Ship in scoped releases: Reserve Page, Buyer Dashboard, Onboarding + Timer, Protection Plans, Order Tracking.
4. Measure every release the same way — site analytics, unmoderated and moderated user studies, distressed-support-call volume, post-purchase surveys — and feed the results into the next one.

Evidence map

Before any screen changed, the diagnosis came from four channels nobody had read side by side. This table is what each one told us, and which release it informed.

Current state vs. future state

The starting point wasn’t a bad page. It was a process that went offline the moment it got hard. The redesign target was a single online spine from reservation to delivery, with the advisor call demoted from a gate to a review.

By 2022 a used-car shopper had three real alternatives to Shift: Carvana and Vroom, both fully online and national; CarMax, which had bolted a buy-online path onto 200+ stores; and the local dealer’s website, which is still a lead form that ends on the sales floor.

On inventory and reach, Shift lost. Carvana and CarMax simply had more cars in more places, and I never designed around that. Where Shift could win was the moment of commitment — and that’s exactly where every competitor had a weakness a buyer could feel:

• Vroom asked for a deposit to hold the car and then went quiet; delivery and title delays were the most common complaint about the brand.
• Carvana ran a clean online purchase, but the only human was a support line, and coverage was an add-on rather than an explanation.
• CarMax handed you to a store to finish, which meant the pace was theirs.
• Dealer sites promised a callback and delivered a salesperson.

So the reservation-based checkout was a bet on three things the matrix makes visible: no deposit to start, a clear picture of what happens next, and a human who reviews instead of sells. The return guarantee and tracking were table stakes among the online players; the onboarding, task dashboard, and Protection Plans task were not.

The research didn’t produce formal persona documents — it produced patterns, and each one mapped to a specific way we were losing a sale. Three composites carry those patterns through the rest of this page:

• Maya, 29 — the cautious first-timer. Won’t put $250 down before she understands the process, and is afraid the car will be sold out from under her. Lost at the deposit page.
• Darius, 41 — the financing buyer. Longest task list, does it in three sittings, loses momentum with no sense of progress or timing. Lost mid-dashboard.
• Priya, 36 — the decisive self-server. Skips anything that reads as an upsell, then spends fulfillment week calling to ask where the car is. Lost on revenue and on support cost.

Shift was asking too much, too early. A $250 non-refundable deposit just to begin checkout, wrapped in a page with thin information, a pushy tone, and low transparency. The stated goal was to filter out bad leads. It filtered out good ones too.

CX data surfaced a second problem hiding underneath the first: customers were afraid the car would be sold out from under them while they finished. The deposit wasn’t calming that fear — it was just charging for it.

Old: “Buy this Tesla today!” — billing info, card number, and a “Place $250 deposit” button before you’d learned anything about how the purchase actually worked.

New: “Ready, set, yours.” — a reservation. Enter contact info and how you want to get the car, and we hold it for one hour while you complete the next steps. No deposit. Trustpilot rating, what every purchase includes, and a plain tooltip explaining exactly what happens if the hold expires.

We solved both problems at once: eliminate the deposit, reframe the page as a reservation. High-intent purchase initiations jumped and bounce rate was decimated — measured through analytics and customer feedback via Sales.

This was the trickiest thing on the roadmap to get right. Removing the deposit removed the commitment signal — the thing telling us who was serious and telling the customer the car was theirs. The one-hour reservation had to do that job, and a countdown clock on a $30,000 purchase is a dark pattern by default.

Testing made the gap obvious: even with the new dashboard, customers were still unsure what to expect and lost steam partway through their tasks. Purchase times exploded and advisors had to intervene.

The fix was to make the timer the least important part of the timer system. Four onboarding screens do the level-setting before the clock ever appears:

1. Complete Your Tasks — we hold your car for one hour while you work, and these tasks finish well inside that window. (Longer if you’re financing with Shift — the constraint should never punish the customer for the path that takes longer.)
2. Review The Details — once you finish, an advisor calls to verify documents and payment and walk through next steps.
3. Time to Sign — payments, e-signature, DMV docs.
4. Hit The Road! — ready?

Then the timer itself: a countdown component in the header, an animation spec, and a tooltip that says exactly what happens when time runs out — we release the hold and refund, other shoppers get the chance, and you can still buy the car if it’s available. No trapdoor.

Completion rate went from 44% to 84%, and low-intent users dropped 12% — the deposit’s real job, done by clarity instead of a credit card.

Protection plans accounted for 50% of Shift’s revenue, and attach during checkout was the key to reversing a −5% QoQ decline. The old task was leaving most of it on the table: vague labeling and a disjointed subhead on the dashboard card, ambiguous content with broken coverage links, only two of three products shown, static numbers that could differ wildly from actual customer rates, and unfocused, inconsistent CTAs.

The design problem was to sell more coverage while being less like the dealership finance office. So the new task educates first and sells second:

• Onboarding screens to set the scene — “Shift the way you think about protecting your car.” A few dollars a month vs. a sudden repair bill; the most you’ll pay at the shop is a small deductible.
• A brief questionnaire about driving habits — how long you keep a car, miles per year, kind of roads — and a short “customizing your coverage offers” moment.
• Personalized plan offerings — Gap Coverage, Vehicle Protection Plan (Premium / Plus / Standard), and a Convenience Package, each with a sample brochure link, plain-language “what this covers,” honest estimates, and an explicit “continue without coverage” option on every screen.
• A more nuanced version embedded in the financing application, where the numbers are real.

The combined effort with cross-functional teams increased gross profit 43% YoY for two years and boosted F&I sales by 200% — measured through self-attach rates, advisor-attach rates, customer feedback via Sales, and site analytics.

The old dashboard offered no way to track your purchase, which is how “dude, where’s my car?” became a high-volume support category. The new order tracking view shows where the car is at every stage.

I partnered with the Fulfillment team to develop status labels and descriptions that were meaningful to customers, consistent with internal fulfillment statuses, and aligned with UX best practices (Baymard, NN/g). Each state — Quality Inspection in Progress, Preparing for Shipping, In Transit with an expected delivery date, Car Delivered — pairs a map, a timestamped shipping timeline, and a “your action items are complete” reassurance block so the customer never wonders whether the ball is in their court.

Result: a significant reduction in support phone calls and emails, measured through customer support data, user testing studies, and post-purchase surveys. In post-purchase testing, 92% of customers agreed or strongly agreed they felt confident in the status of their vehicle order (48 responses; 54% agree, 38% strongly agree, 8% neutral, 0% disagree).

Outcome. I’m proud of the many initiatives I led to reach this point. And just as we turned our gaze to the road ahead, a new behemoth was rushing into view. Under the hood, the platform had been built to reconfigure rapidly in the event of a full-scale pivot — which then happened twice: omni-channel sales, then a marketplace model. The checkout was held.