Lesson 43 — Adding Evidence and Investigation Notes
The investigation is not finished just because you understand what happened.
Can another analyst follow your evidence, challenge your reasoning and continue the case?

What you will learn
Document the investigation so your findings survive a shift change.
Learning objectives
- Write clear, chronological incident notes.
- Record evidence sources, timestamps, entities and query context.
- Separate observations, inferences and unresolved questions.
- Document analyst decisions and response actions.
- Prepare a reproducible investigation handover.
Why this lesson follows bookmarks
Lesson 42 preserved important hunting findings. This lesson explains how to turn those findings, alerts, KQL results and entity pivots into a coherent case narrative.
A saved result is evidence. A note explains its relevance.
Five questions every investigation note should answer
Keep the source of each conclusion visible. A reader should not have to guess which alert, bookmark or log row supports your statement.
Evidence is not interpretation
Observed: SigninLogs recorded a successful sign-in for alex@contoso.com at 01:57 UTC from 203.0.113.50.
Interpretation: This sign-in may be connected to the mailbox activity detected at 02:08 UTC.
Do not turn a hypothesis into a fact
“The attacker signed in at 01:57” may be too strong without corroboration.
“A successful sign-in preceded the mailbox alert; attribution remains under investigation” is more defensible.
The Agent Foskett evidence record
| Field | What to capture |
|---|---|
| Time | Event timestamp, timezone and when the analyst observed it if relevant |
| Source | Table, alert, bookmark, product or other evidence location |
| Entity | User, device, IP, application or resource |
| Observation | What the telemetry actually shows |
| Reproduction | Query, filters, time window and result identifier where available |
| Assessment | Relevance, confidence and competing explanations |
| Next step | Question or action still outstanding |
Example — a useful incident note
Example values are illustrative, not a real incident.
Use incident comments purposefully
Incident comments provide a place to record findings, decisions and handover context. Use clear headings and concise entries; do not paste an entire unfiltered log export into a comment.
Know the difference: comments and activity history
Comments explain what the analyst learned or decided. The incident activity history records changes and actions associated with the case. Both are useful, but neither replaces source telemetry.
Use KQL to make findings reproducible
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
SigninLogs
| where UserPrincipalName =~ "alex@contoso.com"
| where TimeGenerated between (
datetime(2026-10-07 01:30:00) ..
datetime(2026-10-07 02:30:00)
)
| project TimeGenerated, UserPrincipalName, IPAddress,
AppDisplayName, ResultType, ConditionalAccessStatus
| order by TimeGenerated ascRecord the query and time window used, then identify the specific rows that support the note. Avoid assuming that a later rerun will return identical data after retention or ingestion changes.
Capture negative findings carefully
“No related process events found in the queried table and time window” is useful.
“No other malicious activity occurred” is not supported by that limited search.
Preserve uncertainty
Label findings as confirmed, suspected or unresolved. Explain what additional evidence would change your assessment.
Never use a severity label as a substitute for reasoning.
Record the investigation decision trail
The investigation timeline explains environmental activity; the decision trail explains what the SOC did in response. Keep those concepts distinct.
Record containment decisions
If an account was disabled, a device isolated or a session revoked, document the reason, authorisation, execution time, scope and observed outcome.
Protect sensitive information
Use approved case systems and access controls. Avoid copying credentials, tokens, unnecessary personal data or secrets into incident comments.
Handover — what the next analyst needs
| Include | Purpose |
|---|---|
| Incident status and priority | What needs attention now |
| Confirmed findings | What the evidence establishes |
| Entities and scope | What has been investigated |
| Key timestamps and evidence links | How to reproduce the story |
| Actions already taken | Avoid duplicate or conflicting response |
| Unanswered questions and next steps | Continue without restarting |
Agent Foskett exercise
Your incident contains a suspicious sign-in at 01:57, mailbox activity at 02:08 and PowerShell execution on DEVICE-27 at 02:14.
- Write one confirmed evidence statement for the sign-in.
- Write one hypothesis linking the sign-in and mailbox alert.
- Identify which query, time window and entities must be recorded.
- Document a decision to escalate, including the reason.
- Write a three-sentence handover that makes the next investigation step obvious.
Common mistakes
- Writing “malicious” without showing supporting evidence.
- Omitting timestamps or timezones.
- Recording conclusions but not source queries.
- Mixing case actions with attacker activity.
- Leaving the next analyst without a clear question.
Agent Foskett takeaway
If the next analyst cannot reconstruct your reasoning, the investigation is not properly documented.
Capture the evidence. State the uncertainty. Explain the decision. Leave a trail somebody else can follow.
