XCelerate Auto · Product design, 2023-2026
XCare sells extended coverage for electric vehicles. I designed how people buy it and how the team behind it handles the claims that follow.
Context
XCare covers electric vehicles after the manufacturer warranty runs out. Battery, drive unit, drivetrain, electronics, the expensive parts. A plan costs between $2,000 and $5,000, and the customer is usually buying it for a car worth more than they earn in a year.
That price point sets the terms for everything. Nobody buys a $4,000 warranty on impulse, and nobody forgives a company that makes the claim process painful after taking that money. The product had to be convincing at purchase and calm during a claim, which are two different design problems attached to the same business.
Starting point
The company was already good at the actual work. Existing customers were happy, the coverage was real, the support was human. What did not exist was any layer between that and the people who had not bought yet.
Flows went straight from an idea to a developer, sometimes by way of a Miro board. They were live and functional, built from a basic UI kit and plain HTML elements, but nothing had been designed in Figma, tested for conversion, or checked against how competitors handled the same decisions. No design system. No mobile work.
Claims ran on a spreadsheet. One master sheet with every field an adjuster might need, scrolling sideways past the edge of the screen.
Approach
Most warranty companies treat these as separate departments with separate vendors. Marketing owns the site, a sales team owns the quote, an ops vendor owns the claims software, and the customer feels every seam.
I worked on all three, which meant the same language, the same components, and the same status vocabulary run from the ad click to the adjuster's screen. A plan called XCare Premium on the homepage is called XCare Premium in the checkout and XCare Premium in the claim record an adjuster opens two years later.
That continuity is not a flourish. It is the reason a support rep can read a customer's screen over the phone without asking them what they are looking at.
01 · Purchase flow
The extended warranty industry runs on one model: capture the lead, then call it. You enter your details, you get "an agent will contact you shortly," and then you get contacted shortly, repeatedly, for weeks. The category has a bad reputation and this is most of why.
We went the other way. Enter your year, make, model and mileage, and real prices appear on the next screen. No email, no phone number, nothing standing between a person and the number they came for.
The reasoning was about trust, not friction. Someone about to spend four thousand dollars on a promise needs evidence early that the company is not playing games. Showing the price with nothing asked in return is the cheapest possible proof that the pricing is fair and the coverage is what it says it is. You earn that at the top of the funnel so it is already established when the money moves.
We still run gated capture on some paid lead gen campaigns, where the job of the page is different. The ungated quote is the product's front door, not a rule applied everywhere.
The rest of the flow follows from that. Five steps, visible from step one: select plan, owner information, car verification, review, payment. Nothing hidden, no surprise step six.
Coverage type (full Premium versus battery and drive unit only) is a different product with a different risk profile. Deductible is a dial inside whichever product you picked. Those two were deliberately separated and ordered, because a customer tuning a deductible before understanding what is covered is optimizing the wrong thing.
Six coverage terms shown at once, each with its end date, mileage cap, and price. Warranty length is inherently a comparison decision, and anything that forces you to hold one option in memory while looking at the next adds friction to a decision that already has real money attached.
One-time payment or eleven monthly installments at zero interest. For a $3,000 purchase, "can I pay this monthly" changes which plan a person even considers, so it comes before the plans rather than at checkout.
Full plan detail lives in an overlay a person opens when they want it, so the main flow stays a decision instead of turning into a brochure.
02 · Claims, both ends
Everything sold in the flow above eventually shows up here.
A claim is not a support ticket. It runs for weeks, moves between the customer, a service center, and an adjuster, and stalls constantly for reasons that are nobody's fault: parts on backorder, a diagnosis pending, a customer who has not sent photos. The design problem is not "display claims." It is "make it obvious which of the open claims needs a human right now."
The clearest evidence of what was broken came from the adjusters themselves. The master spreadsheet held everything, which meant it was unusable in the two situations that matter most: mid-call with a customer, or mid-claim with a service center. So adjusters quietly built their own private sheets with fewer columns, just enough to work fast. Half the team was operating on a personal copy of the truth.
That is the actual brief. Not "replace the spreadsheet," but "build the stripped-down view they kept recreating by hand, and make it the real one."
Under Review, Needs More Info, Escalated, Service and Diagnosis, Authorized. Five states, mapped to who is currently blocking rather than to how far along the claim is. A claim in Needs More Info is waiting on the customer. One in Escalated is waiting on us. That distinction is what makes the list actionable instead of merely informative.
Real adjuster notes are short, specific, and non-standard: parts on backorder with an estimated arrival, could not replicate the fault, awaiting photos from the claimant. Forcing that into structured fields would have failed, so notes stayed free text and stayed visible in the list, doing the work of a status the system will never have a field for.
A small indicator on the claim ID, so "this customer messaged you and nobody replied" is visible without opening anything. In a queue built around waiting, the most expensive failure is a claim quietly waiting on you.
Vehicle and VIN, adjuster, coverage plan, deductible, and policy dates sit alongside the message thread with the customer. An adjuster answering "is this covered" should not have to leave the conversation to find out.
A policy has a life beyond claims. Owners add a second driver, sell the car and transfer the coverage, or cancel outright. The marketing site sells transferable coverage as a real advantage over legacy providers, and a promise like that has to be carried out by someone on an ordinary Tuesday. Those actions sit in the same customer record as the claim history, so an adjuster taking a call about a transfer is not opening a second tool to do it.
The customer sees this claim too, from their own side. They start it on their phone, then follow it in their portal: which vehicle, what state it is in, what has been paid, with invoices and health reports as PDFs they can pull themselves instead of emailing someone.
Signing in is keyed to a VIN or contract number rather than an email and password, because the thing a person can find when their car has stopped working is the car, not an account they created two years ago.
The status vocabulary is deliberately identical on both screens. When an adjuster says a claim is under review, that is the exact phrase the customer is already reading. Support calls get shorter when nobody has to translate between two systems that describe the same thing differently.
The payment card. The usual warranty pattern is that the customer pays the repair shop, files for reimbursement, and waits to be made whole on someone else's schedule. A virtual card issued to the customer lets XCare pay the service center directly, which removes the part of a claim where a person is out of pocket for work they already paid for once. That is a decision about how money moves through a claim rather than about screens, and it came out of sitting with the people handling the claims.
Both the admin portal and the client portal are in development now, and the design continues alongside them rather than ending at handoff.
03 · Marketing site
EV warranty is a genuinely confusing purchase. Most people do not know that many traditional providers exclude the battery, which is the most expensive component in the car and the entire reason they are shopping.
The site closes that gap fast. The quote calculator sits in the hero, so the primary action is available before any argument is made. Two plans are compared side by side rather than sequentially. A comparison table addresses the exclusion problem directly. The claims process is laid out in four steps on the marketing site, before purchase, because how a warranty behaves when you claim on it is the actual product.
The buying process is spelled out on the site too, ahead of the flow itself. Three steps, and the promise that there is no email capture and no sales call, stated before a person commits to starting. The ungated quote only works as a trust signal if people know it is there.
Campaign landing pages reuse the same components and the same plan language, so a person arriving from an ad and a person arriving from search end up in the same vocabulary.
Working model
Two and a half years of continuous work, still ongoing, in their tools and on their timeline.
Small team: me, one developer, one PM, with the CEO, COO, and the adjusters in the room for the decisions that affected them. That size is why the three surfaces are coherent. A studio would have scoped the site, shipped it, and left before anyone discovered that the plan names in the checkout did not match the plan names in the claims database. Staying meant decisions compounded instead of resetting.
It also meant losing arguments sometimes. Early on I pushed back on how much brand red the plan selection screen was carrying. The COO wanted it, we shipped it his way, and I let it go. Six months later, looking at it fresh, I realized my original argument had been the weak one. It was never about the color being loud. It was that red was doing six jobs at once on that screen, and the filters had ended up visually louder than the buy buttons, which is a conversion problem rather than a taste problem. That is a better argument and I did not have it at the time. It is not near the top of their list right now, so the screen still ships as it is.
Next
The plan selection screen gives six coverage terms equal weight with no recommended option, leaving a $2,000 spread entirely to the customer's judgment. The deductible selector changes the price but does not yet show what it changes it by. The two coverage plans are compared one at a time when the decision is inherently side by side. All three are conversion work, all three are measurable, and all three are on the list.
I work with software teams across product UX, internal platforms, marketing websites, and the space between them.
