How Lens gets its figures — and how to challenge them.

Interviews first. Operators, investors, creators, ad tech, payment providers. Conversations are off the record by default; what gets published is re-verified before it goes to print. It's slow. That's what makes Lens hard to reproduce.

Proprietary scrapes next. App stores tracked weekly, catalogs monitored by country, editorial capsules indexed with their first-appearance date. The scripts live internally; archived snapshots go back to March 2024 for the vertical drama section, March 2025 for Africa.

Operator dashboards, when an NDA contract allows it. We never publish raw data — we publish derived conclusions, and the raw version stays with the operator who shared it.

Lived fieldwork, finally. Ex-CTO Afrostream, Chief Platform Officer, TRACE, a decade of pushing OTT distribution from Abidjan, Lagos, Johannesburg. This isn't desk research. That's why a mobile-money figure that smells wrong is recognizable at ten meters.

Research firms — iiMedia, Sensor Tower, Omdia, Ampere, Data.ai, Global Market Insights, H2 Gambling Capital — when we buy access or explicitly source the cited edition. Every secondary stat carries the firm name and publication date in the caption. Not "according to market estimates." The name, the date, visible to the naked eye.

Public data — World Bank Global Findex, GSMA, UN DESA, African Development Bank, central banks. Rarely spectacular, always defensible.

Named reporting — FT, Bloomberg, The Information, Rest of World, Semafor, TechCrunch, Techpoint Africa, 36Kr. Scoops are cited with link and date, not with "according to press."

Public documents — 10-K filings, prospectuses, regulatory decisions, corporate press releases. Under-used in the streaming industry; often the most reliable source when it exists.

A figure enters Lens when at least two independent sources converge within ±15%, or when a single primary source makes its calculation method and margin of error explicit. Outside that, it's a Lens estimate — and it's labeled as such.

Disagreements between sources are published, not crushed. For 2025, Omdia puts the world vertical drama market at $11B; iiMedia's China forecast (CNY 67.79B, ≈ $9.4B) plus Media Partners Asia's figure for the market outside China ($2.7B) gives $12.1B. Lens doesn't give you the average as if it meant something. We publish both, explain why they diverge — here, the two sources do not draw the same boundary around the market — decide if evidence exists, and let the reader decide otherwise.

Yuan figures converted at the annual average rate (Federal Reserve G.5A: 7.1957 CNY/USD in 2024, 7.1875 in 2025; Q1 2026: 6.93, OECD).

If a figure on this site is wrong, write to me. ludovic@streaming-radar.com. Send the source, the date, the method. Reply within 72 hours.

If you're right, the figure is updated and you're cited inline — permanently, with your name linked to your LinkedIn profile or site. That's the policy. Not PR. That's how the product stays credible three years after launch.

If it's an opinion, we discuss it. If it's a troll, you don't interest me. If you have a better alternative method, I owe you a beer next time we meet.

No scraping behind paywalls. No NDA data shared publicly, even aggregated. No "industry estimates" without verifiable method. No AI-generated prose in published reports — AI tools are used for analysis, never for final writing. No undated figures; an undated figure is a dead figure.

No paid interviews. No product placement. No commissioned reports. Operators who license the reports buy them at their public prices; none has editorial oversight, none can retract a criticism directed at them. It's written into every license contract.

Each report carries a semantic version (e.g. v4.6.0) visible in its header. Minor corrections bump the patch; substantial additions bump the minor; a new chapter or thesis revision bumps the major. The changelog lives in the report itself — final page — and each entry names who flagged the error when applicable.

On the site, every public page exposes its last modification date (dateModified) in metadata and visibly in the footer. Datapoint-driven figures are synced twice a week, Tuesday and Friday at 7h UTC, against the central database — a variance above 2%, or a source change, triggers an automatic rebuild.