FOR INVESTIGATORS
Operational guidance for field investigators, team leads, and anyone coordinating observation.
The single biggest determinant of whether a Hermes case holds up to analysis is the quality of the initial observation record. Everything else automated screening, pattern detection, cohort analysis depends on the geometric and temporal fidelity of what was filed. Your job, as an investigator, is to make that fidelity as high as possible.
How Hermes teams work
A team in Hermes is a coordinated group of observers tied to an area of operations (AO). Teams support three modes of activity:
- Active pings a team member signals "I am observing now, in this direction, at this elevation." Other members in the AO are notified and can corroborate.
- Passive alerts any new report filed inside a team's AO notifies the team automatically.
- Auto-linking when two or more reports arrive within 30 minutes and 50 miles of each other, Hermes links them as candidate corroborating observations and flags the cluster for review.
Team features are under the Teams section. A team's AO is a geofence; members submit using the normal form, but their reports are tagged with the team ID.
How to file a high-signal report
Most intake forms are filled out from memory. A report filed by a trained investigator should not be. Treat the form as a structured interview protocol.
Location precision
- Use lat/lon to 4+ decimal places where possible (~11m precision).
- If the observer moved during the sighting, record the initial position. Note the movement in the description.
- Observer elevation matters; get it right. A handheld GPS or phone altimeter is fine.
Time precision
- Record the start time as the moment the object was first observed, not the moment the observer noticed it was unusual.
- Record the timezone explicitly. UTC offsets are preferred over zone abbreviations (PST, EST) because they resolve unambiguously.
- If the witness has a timestamp from a phone photo or a text message they sent, use that. It is more reliable than their memory.
Geometry capture
- Bearing: point at the object's initial position, or the midpoint of its track. Use a compass (magnetic-to-true correction if necessary) or a phone compass app. A rough bearing with a wide error bar is better than no bearing at all say "120 30" in the description.
- Elevation angle: fist-at-arm's-length is about 10. Four fists is 40. Directly overhead is 90. Be conservative witnesses routinely overestimate elevation for dramatic objects.
- Duration: if you weren't counting, estimate conservatively. "About 30 seconds" is more useful than "several minutes" if the latter is a guess.
The narrative description
The free-text field is not scored by the algorithm, but it is the richest evidence a reviewer will read. Structure it:
- What the observer was doing when the sighting began
- First impression what they initially thought it was
- Physical description, ordered: shape, size (relative to something familiar), color, light characteristics
- Motion: path, speed, acceleration, any direction changes
- Sound or absence of sound
- How the sighting ended faded, vanished, moved out of view, witness looked away
- Witness emotional state, and whether anyone else saw it
What evidence carries the most analytical weight
Ranked roughly in order of how much an additional piece of evidence changes the analysis:
| Evidence | Analytical weight | Notes |
|---|---|---|
| Independent witnesses with non-overlapping viewpoints | Very high | Two people standing in different places reporting the same object at the same bearing+elevation is a geometric triangulation. Two people in the same place reporting the same thing is corroboration, which is less strong but still valuable. |
| Photo or video with preserved EXIF metadata | Very high | Timestamp, GPS, camera model, exposure settings all verify the report. Do not strip metadata. Do not re-encode. |
| Precise bearing and elevation angle | High | Enables celestial and satellite cross-reference. A report without these can't rule out Venus. |
| Instrument readings (compass, inclinometer, range finder) | High | If available, these anchor the witness's geometry to something calibrated. |
| Observations from multiple sensor types (naked eye + IR + camera) | High | Artifacts that appear in one modality but not another are strong diagnostic evidence. |
| Contemporaneous written notes (text messages, voice memos during the event) | Medium-high | Memory degrades fast. A time-stamped contemporaneous artifact is worth more than a careful reconstruction a day later. |
| Weather and celestial conditions | Medium | Hermes retrieves these automatically, but an observer note ("moon wasn't visible, behind clouds") adds useful context. |
| Description alone, no geometry | Low | Still preserved, still archived, but cannot be run through automated elimination. |
Corroboration workflow
When a candidate corroborating report is auto-linked to one of your team's cases:
- Review both reports' geometry. Do the bearings and elevations intersect at a plausible point in space?
- If yes, note the candidate track in both cases' investigator notes.
- If one witness had timing precision and the other didn't, the precise one anchors the event.
- Independent witnesses who didn't know about each other are stronger evidence than members of the same group.
Triangulation from two distant observers is one of the highest-signal outcomes possible in this work. It's rare; when it happens, document it carefully.
What not to do
- Do not tell a witness what they saw. Your job is to elicit their observation, not to shape it.
- Do not file while intoxicated or sleep-deprived, if you're the witness. Note the state honestly.
- Do not edit photos or video before submission. If a frame needs enhancement for analysis, do it as a separate file and keep the original.
- Do not assume a MEDIUM-HIGH confidence grade from the automated pipeline is a conclusion. It is a signal to look more closely, not a verdict.
- Do not post-process your witnesses into your own framing. Their observation, in their own words, is the primary record.
Filing to other networks
When you submit a report, Hermes generates three pre-formatted filings: MUFON, NUFORC, and Enigma. Hermes does not auto-submit to any of them. You and the witness decide which networks the report goes to. The formatted text is on the case page under "Export for Filing."