# Search strategy and release gates

**Market:** English-language US vehicle research. **Product question:** “What do the records say about this model year, and what should I inspect on the actual car?” No traffic or ranking claim is made for the new domain.

## What currently ranks or competes

| Site | Observable page pattern | Lesson for Car Data Index |
|---|---|---|
| [ForCar](https://forcar.org/car-reliability/toyota/avalon/) | Make → model → year comparisons with computed figures and VIN paths | Build useful model hubs and data tools; disclose calculation choices. |
| [CarWhere](https://www.carwhere.com/research/years-to-avoid/methodology) | Year rankings with a published method that adjusts for vehicle age | Raw year counts alone cannot justify a “best/worst” verdict. |
| [CarComplaints](https://www.carcomplaints.com/Honda/Civic/2018/) | Dense model/year/problem navigation and reported owner issues | Users want issue detail and direct routes to a specific year. |
| [VehDB](https://vehdb.com/model-year-scorecard) | A multi-dataset scorecard in a single working screen | A generic complaint count is easy to replicate; add sources and buyer actions. |

These are **features observed on public pages**, not measured keyword traffic. The same NHTSA dataset is used by several competitors, so merely changing the URL or formatting cannot provide a defensible advantage.

## Release path

| Stage | Scope | Measurable gate |
|---|---|---|
| 0. Data proof | Import two years each for Civic, Camry and F-150; inspect mappings and original IDs | Evidence counts reconcile with NHTSA responses; 100% reviewed pages link to the source. |
| 1. Pilot | Publish three fully reviewed model hubs and six to twelve year pages | Google can crawl/render them; sitemap contains only published URLs; actual GA4 organic sessions and GSC page/query data are recorded weekly. |
| 2. Product use | Watch VIN submissions, model-to-year clicks and time spent using year comparisons with privacy-safe aggregate events | People use the working surfaces; avoid judging only by GSC impressions. |
| 3. Expansion | Add a model only where specific year records and a useful inspection checklist can be maintained | Six-week evidence of indexed pages and qualified visits; no template-only expansion. |

The model list in preview mode is a navigation mockup. Its presence is **not** the proposed publication list. Keyword selection needs a live Semrush or comparable export: US volume, trend, SERP competitors, intent, and per-query difficulty. The first research queue is a hypothesis, not a ranking forecast: `2018 honda civic problems`, `2018 honda civic recalls`, `2017 vs 2018 honda civic`, `2018 toyota camry problems`, `ford f-150 2018 recalls`. Review live SERPs and data before committing to these pages.

## Cluster rules

One model hub addresses model-year comparison; a year page handles the evidence for one year. A dedicated “A/C problems,” “transmission,” “recalls” or “years to avoid” branch is added only if it answers a **different** question with enough sourced records and original inspection guidance. Every branch links back to its model and applicable years. Published pages display source check and editorial review dates. If source mappings change, affected pages return to draft.

Do not publish score numbers until there is a defensible denominator such as confirmed US unit sales per model year, age adjustment, coverage analysis and a published method. NHTSA complaints are self-selected reports; recall counts show campaigns, not unresolved repairs for an exact VIN. This avoids the strongest failure mode of a large automated site: confident conclusions from incomplete raw counts.

## Measurement and stop conditions

Track weekly: published URLs, indexed URLs, Google organic sessions by landing page, nonbrand queries and clicks, VIN tool usage, source freshness and inspected source errors. Keep GSC and GA4 as separate signals; investigate discrepancies with server logs and consent/tag implementation. Reassess the pilot after 60–90 days of stable crawlability and publication. No number of pages guarantees Google traffic; if reviewed pages attract no relevant visits, revisit product differentiation and distribution before adding scale.