RESEARCH METHODOLOGY / VERIFY BEFORE PRIORITIZING

Evidence before the recommendation.

Carbide evaluates dealership marketing and search systems by combining technical observation, primary documentation, business context and post-change verification. The method is designed to separate real constraints from noisy reports.

Reviewed by Carbide Digital

ANSWER IN BRIEF

Evidence before the recommendation.

Carbide evaluates dealership marketing and search systems by combining technical observation, primary documentation, business context and post-change verification, in that order. Research starts from the decision that needs an answer rather than from a tool, and the scope records the platforms, templates, departments, vendors and access limits that can affect the result. Evidence comes from several places at once, rendered crawls, server responses, Search Console, analytics, page-speed data, sitemaps, structured data and live search behaviour, because no single crawler is treated as the full truth. Every automated warning is reproduced live before it becomes a recommendation, and expected platform behaviour is separated from defects that actually affect crawling or customers. Findings are ranked by affected demand, confidence, effort, risk and dependency, and each one names who controls the change. Where evidence is incomplete the conclusion is labelled a hypothesis with the next test attached.

Reviewed by Carbide Digital

TABLE

How each kind of evidence is weighted, and what it cannot tell you

A finding is only as strong as the weakest source behind it. A single crawler's warning is never shipped as a recommendation for exactly that reason.

SourceWeightWhat it cannot tell you
Live reproduction on the real siteHighestWhether the behaviour is intentional; that needs the vendor
Search Console and analyticsHighAnything about pages that were never indexed
Server responses and rendered crawlsHighHow a page behaves for a signed-in or personalised visitor
Structured data and sitemapsMediumWhether the markup matches what a reader sees
Third-party audit toolsLow on their ownWhich of its warnings matter at this store
Vendor reportingContext onlyAnything the vendor is being judged on
What we were told in a meetingRecorded, not relied onWhether it is still true

Carbide's research method, reviewed 2026-09-02. Weighting is our own and is applied to our own findings first.

01

#Define the decision

Research begins with the business decision, search opportunity or operational question that needs an answer. Staying anchored to that question keeps a review focused on outcomes such as qualified discovery, index quality, conversion, budget allocation or workflow reliability.

The scope records the platforms, templates, departments, vendors and access limitations that can affect the result.

02

#Collect multiple forms of evidence

A dealership SEO review can combine rendered crawls, server responses, Search Console, analytics, page-speed data, sitemaps, structured data, live search behavior and representative template checks. No single crawler is treated as the full truth.

Marketing analysis can also include channel reports, spend, landing experiences, attribution design, vendor scope and customer-action data where access is provided.

03

#Verify the finding live

Automated warnings are reproduced on the live or approved staging environment before they become recommendations. Expected behavior gets separated from real defects. CMS, inventory, tracking and personalization quirks do not count unless they materially affect crawling, understanding or customer use.

Where evidence is incomplete, the conclusion is labeled as a hypothesis and paired with the next test required to increase confidence.

04

#Prioritize in operating context

Findings are ranked by the value of affected demand, confidence, likely impact, effort, risk and dependency. A systemic template issue can outrank a longer list of isolated page edits, while a high-risk migration defect can require immediate action even before its traffic effect is visible.

Ownership is part of the recommendation. The output identifies whether the dealership, website platform, inventory provider, agency or another partner controls the change.

05

#Validate after release

A recommendation is not complete when it enters a ticket. Verification can include a rendered recrawl, response check, URL inspection, analytics annotation, Search Console trend, performance measurement or monitored question set.

The result informs the next decision and becomes part of the operating history used to evaluate future changes.