Challenge
SEO health for a client-facing website isn’t something you set and forget. A ranking loss, broken indexation, or unnoticed traffic opportunity can cost months of visibility if left undetected. But manual weekly audits are expensive to run and easy to defer: pulling data from Google Search Console, cross-referencing analytics, filtering for signal over noise, and turning findings into actionable decisions is a routine task that requires domain knowledge and burns several hours every week.
Worse, the audit is only as comprehensive as the person running it. A traffic source nobody thought to check, a page with engagement problems, or a redirect chain that crept in unnoticed stays invisible until someone explicitly looks for it. By then, the opportunity cost has compounded.
Solution
84EM built a fully autonomous SEO and traffic analysis system that runs weekly as a set-and-forget automated process. The system operates on a divide-and-conquer principle: Google Search Console answers “are we ranking and found” (technical issues, position changes), while Simple Analytics answers “where do visitors come from and what do they do” (traffic sources, engagement, conversions). Each data source is the authoritative backbone for its own analysis mode; they corroborate each other but never collide.
Every week, the system automatically pulls data from both APIs, runs a comprehensive analysis suite spanning ranking changes, search result visibility problems, content quality decay, indexing status, and opportunity identification, and intelligently files the results back to GitHub: PRs for proven technical fixes (every one requires human review and merge), and GitHub issues for anything requiring a business decision. When the same problem is flagged a second time, the system recognizes it and adds a comment to the existing issue instead of filing a duplicate.
The tool was built for 84em.com’s own operations—the kind of internal infrastructure that allows a solo consultant to ship and monitor without drowning in process—and later extended to operate across a three-site local directory network, proving its scalability across different vertical audiences.
Key Capabilities
- Fully autonomous weekly analysis. Runs on a schedule with zero human setup per cycle. No manual data pulls, no filtering, no summary writing. Every finding is derived from live data, and state is tracked locally so dedupe never misses an issue.
- Dual-analysis separation. Search-focused findings (rankings, technical SEO, indexing) are kept disjoint from traffic findings (sources, device split, engagement quality) with separate dedupe states and GitHub label families. The same issue never gets filed twice.
- Safety-first human approval. Technical fixes open PRs that wait for human review before merging; traffic findings and strategic decisions file as issues that always require a decision. The system never auto-commits or forces a change.
- Scalable across multiple properties. Built for single-site analysis but extended to handle multi-site runs in parallel, each with its own dedupe state and issue namespacing. One cron job reviews three different directory sites in a single session.
Results
Ranking and indexing problems are caught in days, not weeks. Before the system, a loss of position on an important query might go unnoticed for many weeks (or more). Now it’s flagged in the next Monday morning’s automated report. A broken page that returns 404 no longer sits in the index undetected; the system catches it, files an issue, and suggests a solution.
High-value opportunities surface automatically. The system identifies keywords with real search volume that the site ranks poorly for (position 4–15 where a title or snippet improvement could move the needle) and differentiates them from pure content gaps. Instead of waiting for quarterly planning, the team sees and acts on opportunities weekly.
Weekly audits are now a byproduct of normal operations. What used to require an hour or two of focused time every week is now a background process. Every finding that arrives is already filtered against data, deduplicated against live GitHub, and prioritized by impact. The team sees only signal, never noise.
Multi-site operations stay synchronized without additional overhead. The same system operates across a network of properties, with each site’s findings separated by namespace. One run covers three different audiences and geographies in the time it would take to manually audit one.
