Queries covered noa venture capital, noavc, A/O PropTech, Gregory Dewerpe, complaints, criticism, Glassdoor, Trustpilot, indexed LinkedIn posts, X and Instagram discovery. Primary company pages and matched public post URLs were opened. No paid listening tool, private group, logged-in social feed or native analytics was accessed.

Search results and platform pages showed conflicting follower counts and review totals. The report uses the named opened-page observations and exposes freshness. It does not infer growth from cached differences. Comment previews are incomplete, and no comprehensive independent media archive was reviewed.

Reactions combine LinkedIn reaction types; they are not a like-only count. All five selected LinkedIn reaction totals were checked directly in the browser on 22 October 2026. Numeric comment totals were exposed for three posts; the other two remain unknown. The posts span different historical periods and are not comparable reporting-window performance. Related-post counters and comment reaction counts were excluded.

Relative post ages are retained as rendered by cached sources; they are not precise publication dates. All sources were checked on 22 October 2026; a crawl-age label describes the search provider’s returned copy, not a new platform capture.

1. Fix a reporting window, such as the latest 90 days, and obtain complete native post exports for noa and a consistent peer set.

2. Capture reactions by type, comments, reposts, publication dates, impressions and follower snapshots. Rank most-liked posts using like-only counts; report the retrieval date and full denominator.

3. Build a deduplicated, identity-checked listening corpus. Separate owned, employee, founder, investor and unaffiliated commentary. Label sentiment manually or validate automated labels against human review.

4. Calculate engagement per impression as (reactions + comments + reposts) / impressions × 100 only when the period and numerator match. Disclose sample sizes, deleted content, platform access limits and review recency.

The supplied African Deck(1).pdf informed the monochrome palette, oversized typography and fine rules only. This dashboard’s content and structure were created for this research. No proposal content, client marks or deck imagery were copied.

This is a static, offline-capable report. Source links require internet and may require platform sign-in. There are no analytics trackers, external font requests or embedded social feeds. “Export evidence” downloads the complete curated dataset; post filters change the view, not the export.