Skip to main content
Analysis & reports

Automating an audit: from findings to an action plan

A tool can crawl a website and produce a long list of warnings. For the person receiving the report, the question often remains the same: where should I start? In the analysis tools I am developing, I work as much on reporting as on collection: explaining the finding, showing its basis and preparing the next action.

AI & automationBy Julien BaillyText updated on 3 min estimated read

Technical observations connected to lines in a report, alongside a magnifying glass, in ceramic and mint glass

Key takeaway

A useful report connects each recommendation to an observation, states its limits and explains how to verify the fix. Automation helps with collection; priority also depends on the website's uses and context.

Make every finding verifiable

A warning such as “broken links” leaves plenty of work to repeat. In a sample report, I prefer to find the page containing the link, the target address, response received and check date. The person making the correction can then reproduce the problem.

The report should also state what was not examined: private pages, unavailable access or an interrupted analysis. An absence of data remains missing information. It requires an additional check before drawing a conclusion.

Turn the warning into an action suited to the website

The same symptom may require a change to content, CMS settings, the theme or code. The recommendation should indicate where to look, what to confirm and what result to check. A generic instruction can provide a starting point, but it must be compared with the actual installation.

This is one direction for my SEO toolbox: producing reports with observable evidence and instructions suited to the identified technology. The tool is still in development. Technical and editorial review are part of the process.

Choose priorities using the business context

An essential page becoming inaccessible and wording to improve on a secondary page need different responses. To order the work, I look to clarify the affected use, the scale of the problem, the strength of the diagnosis and the dependencies of the fix.

AI can help prepare a summary of these elements. I then keep a visible distinction between the observed fact, the proposed explanation and the point to confirm. The reader should be able to trace the data supporting the recommendation.

  • Which use of the website is affected?
  • What evidence allows the finding to be reproduced?
  • Who can fix it, then check the result?

Plan verification after the fix

A completed action should be verifiable through the appropriate check. To compare performance measurements, conditions must also be documented: Lighthouse notes that the network, load and environment can change results without a code change.

Keeping observations and dates helps track what has been addressed and what remains to examine. This is also the principle behind the reports and comparisons I am developing around Magento Black Box: providing reference points for the next intervention.

Sources and references

Documentation consulted when writing this guide. Advice and procedures should be adapted to your own context.

Related projects