Qualitative analysis of conflict narratives: scale testimony

The AP/PBS story (published Aug 20, 2025) documents a July 12 attack on Shaq al‑Noum in Kordofan that witnesses and rights monitors say left scores dead and homes burned. The first-hand testimony and satellite imagery cited in the piece are exactly the kind of mixed textual + geospatial evidence qualitative teams need to synthesize quickly and responsibly (source: www.pbs.org/newshour/world/villagers-offer-harrowing-accounts-of-one-of-the-deadliest-attacks-in-sudans-civil-war).

• Why researchers care: attacks like this are time-sensitive evidence for humanitarian triage and accountability.
• What’s hard: dozens of interviews, inconsistent language, PII risks, and contradictory casualty counts.
• Supporting data to consult: UNICEF casualty brief and Yale Humanitarian Research Lab satellite analysis (see www.unicef.org and Yale HRL referenced in the PBS piece).

Plain English: RSF-aligned fighters and allied militias entered Shaq al‑Noum, looted homes, set fire to straw houses, shot people fleeing, and committed sexual violence according to multiple eyewitnesses. Satellite imagery corroborated large fires and destruction on July 13–14.

• Key evidence types in the reporting: survivor interviews, witness quotes, NGO casualty tallies, and satellite imagery.
• Contradictory counts are common: local rights groups reported 'at least 200, ' while UNICEF aggregated a wider two-day toll exceeding 450 for the area.
• Analytic priorities: map temporal sequence (attack → fires → displacement), tag actors (RSF, Janjaweed), catalog harms (killings, sexual violence, looting), and geolocate damage clusters from imagery.

Limitations to note: interviews in the piece are a small sample (5+ villagers cited); casualty estimates vary by scope and geography. Treat textual claims as qualitative evidence to be triangulated, not as finalized counts.

Run this as a 2-week pilot to process 50–200 interview files and a small NGO spreadsheet.

• 1) Collect: centralize audio, transcripts, NGO tallies, and satellite image links into one import folder.
• 2) Transcribe: use Evidano transcription with custom dictionary for local names; apply PII redaction.
• 3) Translate: auto-translate non-English quotes while preserving original text.
• 4) Seed codebook: import a starter codebook (actors, harms, locations) or generate one from the corpus.
• 5) Auto-code + verify: run AI-assisted coding, then triage disagreements for manual review (10–15% sample QA).
• 6) Triangulate: link coded quotes to NGO tallies and satellite timestamps; flag mismatches.
• 7) Visualize: export frequency tables, co-occurrence networks, and a short slide pack with clickable evidence.
• 8) Produce deliverables: protection priority list, accountability evidence package, and a one‑page ops memo.
• 9) Handoff: share role-based dashboards with partners; retain a redacted archival export for investigators.

Ethics note: treat this workflow as research-focused and non-diagnostic. Secure consent, limit access to raw testimonies, and anonymize before sharing.

The Shaq al‑Noum reporting (July 12 attack; article published Aug 20, 2025) shows the value (and the difficulty) of turning scattered survivor testimony into operational evidence.

• If you run field teams or analyze conflict narratives: prioritize a short Evidano pilot to centralize transcripts, apply PII-safe coding, and produce auditable themes within days, not months.
• Start by importing a representative batch of interviews and NGO tallies; run the 9-step workflow above and review outputs with protection focal points.

Ready to try this on your corpus? Start a secure trial at www.evidano.com and see how AI-enabled qualitative research reduces manual coding time and produces reproducible, shareable evidence packages.