Owner: Beau Wheeler, Director of Product Team: Mobile
Onboarding ADESA, Carvana, and Vroom exposed a fundamental gap in our native mobile app, which was built for SMB customers, meaning one person, one record, and one linear flow. Our new enterprise customers run assembly-line facilities where different specialists gather vehiclecondition data before sale, working on different sections of the same vehicle concurrently or sequentially, and the current product supports neither model. This means Damage Tagging no longer meets the needs of our current users and cannot scale to support additional enterprise use.
Only one user can be in a record at a time, and sections must be completed in a fixed order, so when the system blocks them, inspectors do what anyone would-they fabricate entries or mark “no damage” just to move forward. The record looks complete, but the data isn’t accurate. The data flowing into consumer-facing listings and wholesale condition reports is false. This creates financial liability and reputational risk at every step downstream. When those records are wrong, sellers face returns, disputes, and out-of-pocket costs on $\$ 20,000+$ transactions, eroding buyer trust and wiping out dealer margins at scale.
Damage Tagging requires a full rebuild. Overall, this work is a continuation of the overall ‘Remove the Duct Tape’ initiative in the mobile app to meet the needs of our enterprise customers while also addressing product and tech debt. The existing architecture cannot be extended to support concurrent multi-user workflows because it lacks a concept of independent sections. The entire flow, navigation, and state model assumes linear progression.
Workflow and Interface: The new workflow removes the constraints that break enterprise operations. Sections can be worked in parallel or in sequence. Multiple inspectors can own different parts of the same vehicle simultaneously. No single user owns the whole record; it comes together as each inspector completes their section. The system adapts to how a facility runs. The interface is redesigned around that model, optimized for field speed, with no blocked steps and no need to fabricate entries.
Data Quality: The data experience is simplified alongside it. The taxonomy already exists, but is too complex to use accurately in the field. Too many choices with no hierarchy means inspectors default to whatever is fastest, not whatever is accurate. The rebuild would surface the most-used damage types first, reduce friction on native mobile, and guide inspectors to the right entry rather than the nearest one.
Getting this right now is not just a one-off product fix. Accurate, validated damage data captured at scale is the prerequisite for additional products, such as future in-app condition reports and Al-powered damage detection, both of which cannot be built based on fabricated entries and bad data. Without this rebuild, we cannot reliably serve ADESA, Carvana, or Vroom at scale, and the data pipeline for future products remains compromised.
Success will be measured by data accuracy against condition reports and reduction in placeholder submissions. Baseline established at launch; targets measured at 60 and 90 days post-release.
| Metric | Baseline | Target (60 days) | Target (90 days) |
|---|---|---|---|
| Section accuracy vs. condition reports | 58% | 72% | 85% |
| Placeholder “no damage” rate | 34% | 18% | <8% |
| Records requiring post-publish correction | 20% | 11% | <5% |
| Time from inspection start to publish | 47 min | 35 min | 28 min |