Methodology

We pull the complete political-ad archive for each monitored country and re-pull daily. Every observation is stored append-only: one record per ad per pull. This matters because the archive itself is unstable — ads and entire advertiser pages can disappear between pulls. Append-only snapshots let us report what was visible at each point in time, and preserve records the archive no longer shows.

Three model-assisted layers run over the raw archive, each validated against a hand-labeled evaluation set before deployment, with accuracy measured per task:

• Triage(is this ad domestic politics, foreign spillover, or miscategorized commercial content?) — 98% measured accuracy. Roughly 87% of ads "reaching" Zambia are not domestic politics; raw Ad Library counts overstate domestic activity severalfold.
• Actor attribution (which registered political actor is behind this page?) — 99% measured accuracy against a curated registry, with abstention when evidence is insufficient. Supporter pages praising a candidate are not attributed to that candidate.
• Message analysis (topics, tone, call to action, language) — 86% measured accuracy across a 13-topic taxonomy.

Classification runs at the level of creative families (identical text from the same page), and family keys are page-scoped so that Meta's placeholder texts can never propagate a classification across unrelated advertisers.

• Meta withholds creative content for ads that ran without required disclaimers or were removed — in Zambia, the large majority of domestic political ads. Message analysis covers only the visible slice, which skews toward compliant, established advertisers.
• Spend and impressions are reported by Meta as ranges; we use midpoints and mark all such figures as estimates.
• The archive covers Meta platforms only — not TikTok, YouTube, radio, print, or organic content.
• Coordination detection can only text-match ads whose creative survives; a null result is bounded by that coverage.

Questions, corrections, or data requests: reach out via GitHub (jknap/ballotscope) — or the contact details published alongside our reports.