Appendix 6 — SPES Data Inputs & Outputs Schema

(Decision-Grade Information Architecture for Public Engagement)

Purpose of this appendix
To define—precisely and non-aspirationally:

• what data a SPES must actually ingest,
• where that data comes from,
• how often it updates,
• what decisions it supports,
• and what it explicitly does not attempt to do.

This appendix exists to prevent two institutional failures:

1. treating SPES as a “communications generalist,” and
2. drowning the role in data that does not translate into decisions.

6.1 Design Principle (Non-Negotiable)

SPES does not collect data for insight.
SPES collects data for decision constraint.

Every input must:

• change a choice,
• bound a claim,
• trigger escalation,
• or justify restraint.

If it does none of those, it is noise.

6.2 Core Input Domains (What SPES Must Track)

SPES data inputs fall into six domains, each mapped to a specific decision function.

Domain A — Embassy Public Outputs (Authoritative Voice)

What this includes

• Press releases
• Media notes
• Social media posts (FB, X, IG, YouTube)
• Ambassador / DCM speeches
• Op-eds
• Official statements

Granularity

• Full text
• Timestamp
• Platform
• Clearance path (if known)
• Whether later corrected/clarified

Update cadence

• Continuous (daily ingestion)

Why it matters

• These outputs are treated as policy by external audiences.
• SPES must know exactly what has already been said.

Decisions supported

• Can we say this again?
• Are we contradicting ourselves?
• Has the Embassy already implied a position?

What SPES does not do

• Rewrite after publication
• Judge tone aesthetically
• Attribute blame for errors

Domain B — Engagement & Interaction Data (Risk Surface)

What this includes

• Post engagement metrics (likes, shares, comments)
• Comment content (sampled)
• Escalation patterns
• Media pickup of social posts

Sources

• Meta Business Suite
• X analytics
• Manual media scans

Granularity

• Post-level, not individual-level
• Aggregate patterns, not profiling

Update cadence

• Weekly review
• Real-time flagging for spikes

Why it matters

• Engagement reveals where content is being reinterpreted.
• Comment threads show when PD becomes grievance intake.

Decisions supported

• Respond / hold / disengage
• Escalate to Front Office
• Adjust future framing

Explicit exclusions

• Individual demographic profiling
• Sentiment scoring for its own sake
• Influencer targeting

Domain C — Section & Agency Activity Signals (Internal Reality)

What this includes

• Section events (Econ, Pol, DAO, CDC, Consular, etc.)
• Program launches or milestones
• Agreements, MOUs, shipments, visits
• Grants, exchanges, trainings

Sources

• Section calendars
• Weekly staff meetings
• Internal emails
• Front Office updates

Granularity

• Event-level
• Program-level

Update cadence

• Weekly structured intake
• Ad hoc for sensitive activities

Why it matters

• Most comms failures occur when PD is surprised.
• SPES must see activity before messaging pressure hits.

Decisions supported

• What is safe to message
• What should stay quiet
• What requires alignment language

What SPES does not do

• Approve programs
• Set section priorities
• Represent sections externally

Domain D — Policy & Strategy Constraints (Outer Boundaries)

What this includes

• Integrated Country Strategy (ICS)
• NSS / NDS priorities
• State press guidance
• Travel advisories
• Visa policy changes
• Congressional testimony (relevant)

Sources

• State.gov
• Embassy cables / guidance
• Public testimony

Granularity

• Claim-level
• Language-level (how things are framed)

Update cadence

• Event-driven (policy change)
• Quarterly baseline review

Why it matters

• SPES must know when older assumptions are invalid.
• Prevents “ICS ghost logic.”

Decisions supported

• Is this still aligned?
• Are we exceeding mandate?
• What language is now unsafe?

Domain E — Media Environment (External Interpretation Layer)

What this includes

• Local media narratives
• Recurrent outlets/themes
• Politically aligned outlets
• Social amplification patterns

Sources

• Daily press scans
• Media summaries
• PAO briefings
• LES insights

Granularity

• Outlet-level
• Narrative-level (not article-by-article exhaust)

Update cadence

• Daily skim
• Weekly synthesis

Why it matters

• The Embassy does not control meaning—media does.
• SPES tracks how outputs travel.

Decisions supported

• When silence is safer
• When clarification is required
• When holding language is enough

Domain F — Audience Architecture (Structural, Not Targeting)

What this includes

• Known PD audiences:
  • students
  • journalists
  • civil society
  • business community
  • alumni
• Exchange participation data (aggregate)
• American Spaces usage patterns

Sources

• ECA data
• Fulbright/IVLP records (aggregate)
• American Spaces reports
• PD unit inputs

Granularity

• Program-level
• Aggregate trends only

Update cadence

• Quarterly

Why it matters

• Prevents fantasy audiences.
• Anchors PD in reality.

What SPES does not do

• Audience targeting
• Demographic micro-segmentation
• Political profiling

6.3 Output Classes (What SPES Produces)

SPES outputs are decision artifacts, not reports for their own sake.

Output Class 1 — Leadership Snapshots

Audience: Ambassador / DCM
Frequency: As needed
Form: 1 page max

Includes

• What changed
• What’s risky
• What not to say

Output Class 2 — Risk Registers

Audience: PAO / Front Office / Legal
Frequency: Living document

Includes

• Trigger conditions
• Escalation thresholds
• Ownership mapping

Output Class 3 — Narrative Drift Reports

Audience: PAO / Section Chiefs
Frequency: Quarterly

Includes

• Repeated frames
• Silent gaps
• Overexposed themes

Output Class 4 — Holding Language Menus

Audience: PAO / DCM
Frequency: Event-driven

Includes

• What can be said
• What cannot
• What to defer

6.4 Input → Output Mapping (Accountability Layer)

If an input does not map to an output, it should be dropped.

6.5 Common Failure Modes (Explicit)

SPES data systems fail when:

• everything is tracked
• nothing is prioritized
• insights are not tied to decisions
• outputs are descriptive but not constraining

This appendix is designed to prevent that.