#087Guide4 min

Understand what changes between two analyses

Read the previous-scan comparison and distinguish a change in recorded findings from a verified correction.

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.

01Open the comparison for the same project

Open Projects, select the project you want to follow, then use Scan history to open a completed report. When an earlier eligible analysis exists for that project, the report displays Since the previous scan. The card shows Severe and Total counts before and after, with a message such as Same number of severe issues. Use Open previous scan to inspect the earlier report rather than relying on the summary alone.

The earlier report is selected automatically from finished analyses of the same project, using completion time. Both successful and partial analyses can participate if they have a finish time. It is not an arbitrary comparison of two projects, and the card does not offer a baseline picker. If this is the first eligible report, or the current analysis is not finished successfully or partially, there is no eligible pair and the card is not displayed. Its absence is not a finding about the application’s security.

02Read the counters as recorded findings

Total counts the findings recorded for each analysis. Severe counts the technical high and critical findings. The report’s customer-facing severity words are Danger, To fix and To watch; do not assume Severe means only the findings labelled Danger. A high technical finding belongs to To fix and is still included in the comparison’s severe count. The comparison is a count of recorded findings, not a count of exploitable attacks or affected people.

For example, a change from five severe findings to three means two fewer severe findings were recorded overall. It does not establish which two were fixed. Three old findings could have disappeared while one new severe issue appeared. Likewise, equal totals can hide a completely different set of findings. The comparison card does not match vulnerabilities one by one, identify code changes or provide a list of newly introduced and resolved issues. Open the underlying reports to investigate those questions.

03Check coverage before claiming progress

Compare the target and the selected checks in both analyses, then read their completion status and available coverage information. A partial analysis may have fewer recorded findings because some work did not complete. A repository analysis and a narrower check do not necessarily cover the same files or behaviour. Unticked checks are not executed or billed, so a reduction after deselecting a check cannot be presented as proof that its previous findings were corrected.

Changes in evidence can also have legitimate explanations other than a regression or fix. A developer may have updated a dependency, removed a public endpoint or changed an access rule. A newly available security advisory may identify a problem in an unchanged package. These are examples to investigate in the report evidence, not diagnoses generated by the comparison card. Record the explanation only after checking the relevant versions, locations, configuration or deployment with the person responsible.

04Turn a numeric change into a reviewable conclusion

For a claimed correction, keep the earlier finding reference, the reviewed change and the follow-up analysis together. Establish that the new analysis examined the intended target after the change and actually included the relevant check. If a fix was prepared by Hackaiz, it is a draft GitHub pull request that a human reviews and approves; it is never merged automatically. A proposal existing is not evidence that the deployed application contains it.

Write a narrow conclusion, such as “the later report records fewer severe findings; the dependency version was updated and rechecked,” only when the evidence supports each part. Avoid converting a lower count into a security percentage or an assurance that no issue remains. Further analysis work uses credits. Optional monitoring is only for connected systems; it is not required to read two existing reports and does not change the limits of this count comparison. Keep partial or missing coverage visible in any progress summary you share.

Frequently asked questions

Was this article helpful?

Related articles