Rebuilding a Fragmented Aftermarket: How Webskyne Delivered a 3-Sided Automotive Salvage Marketplace with AI-Powered Compatibility

# Case Study: Rebuilding a Fragmented Aftermarket with an AI‑Powered Salvage Marketplace ## Overview The U.S. automotive aftermarket is vast, but the salvage and used‑parts segment remains fragmented and inefficient. Salvage yards operate with inconsistent data standards, repair shops struggle to verify part compatibility, and mobile mechanics waste time coordinating installations. A fast‑growing startup approached Webskyne to transform this chaos into a cohesive, scalable marketplace that could connect three user groups: salvage yards (supply), buyers (demand), and mechanics (service). The engagement focused on turning a prototype into a production‑grade platform with AI‑driven compatibility, multi‑tenant inventory ingestion, and mobile workflows built for real‑world field conditions. Over seven months, we delivered a unified system that improved conversion, cut time‑to‑installation, and created a foundation for analytics‑driven growth. ## Challenge The client faced four interconnected challenges: 1. **Inconsistent inventory data:** Salvage yards maintained catalogs in dozens of formats—spreadsheets, legacy dealer management systems, and sometimes handwritten logs digitized into CSVs. Mapping those to a normalized schema was error‑prone and slow. 2. **Low buyer trust:** Buyers often abandoned carts because they could not confirm whether a part would fit their specific vehicle variant. The absence of reliable compatibility data meant high returns and low confidence. 3. **Operational bottlenecks:** Coordinating installations across independent mechanics was chaotic. The business relied on phone calls and manual scheduling, creating delays and missed appointments. 4. **Limited analytics:** The client lacked visibility into core marketplace metrics such as yard activation, inventory quality, conversion rates, job completion, and on‑time performance. These issues threatened the startup’s ability to scale and limited investor confidence. ## Goals Webskyne aligned the program around clear, measurable goals: - **Build a three‑sided marketplace** for yards, buyers, and mechanics with unified account management and role‑based experiences. - **Improve compatibility accuracy** through AI‑assisted parsing and matching, reducing return rates and cart abandonment. - **Accelerate fulfillment** with a mobile‑first mechanic workflow, including geo‑fenced job start and photo‑based proof of work. - **Enable data‑driven operations** with dashboards for activation, conversion, job performance, and revenue metrics. - **Design for scale** using a modular backend and infrastructure that could grow with inventory and traffic demands. ## Approach We structured the engagement across four phases: discovery, system design, iterative delivery, and optimization. Each phase included weekly demos and prioritized feedback loops to keep the client’s business stakeholders and technical team aligned. ### 1) Discovery & Domain Mapping We conducted interviews with yard managers, repair shop owners, and field mechanics to understand the end‑to‑end workflow. The most critical insight was that compatibility data was the single largest lever on trust and conversion. If we could reliably answer “Will this part fit my vehicle?”, we could increase purchases and reduce support overhead. We mapped the domain into three workflows: - **Supply workflow:** yard onboarding → inventory ingestion → quality validation → pricing & promotions. - **Demand workflow:** vehicle search → compatibility validation → checkout → installation booking. - **Service workflow:** mechanic assignment → geo‑fenced job start → completion proof → payout. ### 2) Data Strategy & Compatibility Engine Compatibility data was inconsistent and noisy. We introduced a multi‑stage pipeline: - **Ingestion adapters** for CSV, Excel, and legacy formats. - **Normalization layer** that enforced a canonical part schema. - **AI‑assisted enrichment** to extract make/model/year/trim data from descriptions. - **Compatibility rules engine** that combined AI predictions with OEM reference datasets and historical transaction outcomes. This hybrid approach balanced speed with accuracy, allowing the business to scale while continuously improving precision through feedback. ### 3) Marketplace & Mobile Experience We designed role‑specific interfaces: a yard portal, a buyer marketplace, and two mechanic apps (owner + field tech). The mechanic flow was designed for field reliability: offline‑tolerant, minimal taps, and clear photo evidence capture. ### 4) Analytics & Governance We embedded metrics at every step, providing dashboards for: activation, inventory health, conversion funnel, job acceptance, and on‑time completion. This allowed leadership to identify drop‑offs and fix operational issues quickly. ## Implementation ### Architecture We built the platform on a modular service architecture using a NestJS backend with Postgres. The web experiences used Next.js for SEO‑friendly marketplace pages and a responsive yard portal. The mobile apps were built in Flutter for consistent behavior across iOS and Android. The infrastructure leveraged AWS for scalability and Vercel for web hosting. Key components included: - **Auth & identity:** Cognito + JWT for secure multi‑tenant access. - **Inventory ingestion:** asynchronous job queue processing with validation and enrichment. - **AI compatibility engine:** model‑assisted matching with rule‑based overrides. - **Payments & commissions:** modularized logic for subscription plans, transaction fees, and installation payouts. - **Geo‑fencing:** location validation for mechanic job start to prevent fraud. ### Inventory Ingestion Pipeline We created a pipeline that could ingest 50k+ parts per yard with minimal manual intervention. Each inventory upload ran through validation (required fields, pricing consistency, year ranges), enrichment (part‑type recognition, VIN decoding where available), and a confidence score. Low‑confidence items were flagged for manual review, allowing the system to maintain quality without slowing the overall process. ### Compatibility Engine The compatibility engine blended AI extraction with deterministic rules. We created a “confidence ladder” where matches scored below a threshold required a second signal (OEM dataset confirmation or historical fit data). This significantly reduced false positives while keeping the marketplace responsive. We also embedded feedback loops: whenever a part was returned for incompatibility, the system updated the confidence model and flagged similar listings. Over time, this improved the accuracy of the matching pipeline. ### Buyer Marketplace We redesigned the buyer journey with clarity and trust as priorities. Search results prominently displayed fit confidence and installation availability, while the checkout flow allowed buyers to opt into installation services in a single step. This reduced drop‑offs and encouraged higher‑value orders. ### Mechanic Workflow The mechanic owner app handled scheduling and assignments. Field mechanics used a simplified app that required geo‑fenced check‑in before job start. Photo evidence and a buyer signature were required for completion, dramatically reducing disputes and payout delays. ### Analytics & Reporting We delivered dashboards tailored for each user group: - **Operations dashboard:** conversion rate, order volume, and average order value. - **Yard analytics:** inventory quality score, sales velocity, and demand hotspots. - **Mechanic metrics:** acceptance rates, on‑time completion, and average installation duration. The system allowed leadership to run experiments—such as reduced commission rates or promoted listings—and measure impact within days. ## Results The platform delivered measurable improvements within the first 90 days of launch: - **Higher conversion:** Buyer conversion improved as fit confidence became visible and returns decreased. - **Faster fulfillment:** The average time from purchase to completed installation dropped substantially due to streamlined mechanic scheduling. - **Supplier activation:** New yards could onboard and publish inventory in hours instead of weeks. - **Operational clarity:** The team gained visibility into conversion funnels and could identify weak points quickly. ## Metrics (90‑Day Post‑Launch) - **+28% buyer conversion** in the marketplace funnel. - **‑32% part returns** due to improved compatibility confidence. - **‑40% average time‑to‑installation** from purchase to completion. - **+45% supplier activation rate** in the first month after onboarding. - **+22% average order value** driven by bundled installation services. - **>92% on‑time mechanic completion rate** after geo‑fencing and evidence capture. ## Lessons Learned 1. **Compatibility is the trust engine.** In a fragmented market, accurate fitment data is the difference between a browsing marketplace and a revenue‑generating one. AI helps, but only when paired with deterministic rules and feedback loops. 2. **Operational workflows drive retention.** The mechanic flow was not just a feature—it was a retention system. When mechanics are paid quickly and disputes drop, supply quality improves and fulfillment becomes reliable. 3. **Data quality must be built into ingestion.** Every manual clean‑up is a tax on scale. By enforcing quality checks and automated enrichment from day one, the system avoided future bottlenecks. 4. **Analytics create compounding advantages.** With clear dashboards, the client could run pricing experiments, optimize commissions, and identify top‑performing yards and mechanics. ## What’s Next The client is now expanding into predictive pricing, demand forecasting, and AI‑assisted inventory recommendations. The architecture delivered by Webskyne supports these initiatives without major refactoring, positioning the platform for long‑term scale and investor confidence. --- **Cover Image:** https://images.unsplash.com/photo-1489515217757-5fd1be406fef?auto=format&fit=crop&w=1400&q=80 **Additional Image:** https://images.unsplash.com/photo-1503376780353-7e6692767b70?auto=format&fit=crop&w=1400&q=80