#095Guide4 min

Measure security progress

Read the security score and scan comparisons alongside coverage and verified corrective work, rather than treating a rising number as a guarantee.

Hackaiz team · updated 21/11/2026

Illustration: an engineer watches charts on several screens.
Illustration: an engineer watches charts on several screens. Illustrative photo, not a screenshot.

01Choose a baseline you can explain

Progress needs a reference point: a particular project, a version, and an analysis date. In the project detail page, Scan history lets you reopen the starting report and later runs. Choose a completed or partial report whose scope you understand, and note any missing checks. If you compare a repository analysis with a later live-site analysis, you are also changing the evidence available, not just measuring the effect of your fixes.

For example, a team may repair a secret exposed in application code and then add checks against the deployed service. The later report can contain more findings because the review became broader. That is useful discovery, not automatically a security regression. Describe separately what was corrected, what was newly examined, and what could not be evaluated. These distinctions make a progress update actionable even when the headline number moves in an unexpected direction.

02Understand the Security score panel

The project detail page displays Security score using findings from successful and partial analyses, excluding dismissed findings. It presents the latest score out of 100, a descriptive grade, and, when the value changed, a difference since the previous scan. The trend uses up to twelve recent qualifying runs. With only one qualifying run there is no earlier point from which to establish a direction.

The panel explicitly warns that a high score does not guarantee the absence of vulnerabilities. Partial runs can contribute to the trend, so inspect execution status and coverage before using the curve in a management update. Do not call a rise a measured reduction in every real-world risk. It is an indicator derived from the findings available to these runs. Changes to the tested target, the checks selected, or which observations remain included can affect that indicator.

03Move from a trend to individual findings

Compared with the previous scan on the project page separates findings that appeared from those no longer present. Review the two analysis dates and investigate the underlying evidence. The interface uses Proven fixed for findings absent from the later comparison, but your explanation should still state whether the same relevant check ran. The count difference in Scan history is simpler: it compares totals and cannot identify which corrective action caused a change.

Use Danger, To fix, and To watch to communicate the observed priorities rather than reporting only a total. Clearing a serious exposure and finding several low-priority signals can be meaningful progress even if the count rises. Conversely, hiding an observation from a score is not a software repair. In expert view, a Security progress panel is available when findings exist; its remaining-risk figure weights open detections. Treat it as an indicator of recorded open findings, not a probability of compromise.

  1. 1Open the baseline report and the later report for the same project.
  2. 2Check target identity, execution status, and comparable coverage.
  3. 3Read the changed findings rather than relying on the score delta.
  4. 4Link each claimed correction to the change made and the later verification evidence.

04Write a progress update that supports a decision

A useful update might say: “The exposed credential identified in the earlier version was replaced and removed from the client bundle; the later check did not report that exposure. The new live-service checks also identified an access-control issue that remains open.” This is an example of wording to adapt only when your actual reports support it. It explains a repaired issue and the remaining work without inventing a percentage improvement.

When a GitHub correction is offered, the draft pull request still needs human review and approval; Hackaiz never merges it automatically. A proposed change is not a deployed fix. After the team applies and deploys the approved correction, run the relevant checks again. Analyses consume credits; unticked checks are neither run nor billed. For websites and APIs, domain ownership proof is required before execution. State any untested area and the next review action alongside the progress you can substantiate.

Frequently asked questions

Was this article helpful?

Related articles