Your encoder sends one stream. Ten thousand people watch it, on phones on 4G, on laptops through corporate firewalls, on a boardroom TV. The machinery that makes that possible is the content delivery network and the player, and a few decisions here decide whether "it worked for us in the gallery" also means it worked for the audience.
The CDN
A content delivery network takes the single stream from the platform's ingest point, converts it into small chunks, and copies them to servers around the world so each viewer fetches from somewhere close. YouTube, Vimeo, Microsoft, Zoom and every dedicated event platform run or rent one. You rarely choose the CDN directly, but its behaviour — how many viewers it can serve, how it copes with corporate networks — is part of choosing the platform.
Adaptive bitrate
The platform transcodes your stream into a ladder of qualities (1080p, 720p, 480p, 360p) and the viewer's player switches between them as their connection changes. This is why you send one good 1080p stream and let the platform make the smaller versions — and why sending a bad stream can't be fixed downstream. The rungs of the ladder come from your source; garbage in, garbage in at every size.
Latency: the trade-off you have to choose
Standard HLS delivery is 15–45 seconds behind live — robust, cacheable, works everywhere. Low-latency HLS and similar modes get to 3–8 seconds at some cost in resilience. WebRTC-based delivery reaches under a second for genuine two-way interaction but scales less cheaply. Choose based on what the audience needs to do: watch (standard is fine), react to live polls (low latency helps), converse (you need a video call, not a stream). Don't pay for latency you won't use.
Players and where they live
The player is the bit of software on the viewer's screen. It might be YouTube's, embedded in your platform; the platform's own; or a player embedded on your website. Considerations: does it support captions and accessibility controls; does it work on the devices your audience uses; does it fall back gracefully when the stream drops; and can you brand it. Test it on a phone, a locked-down corporate laptop and a smart TV before you commit.
Corporate networks: the silent killer
Many organisations block YouTube, throttle video, or route everything through a proxy that breaks streaming protocols. If your audience is inside a company, ask their IT team to test the exact player and stream type in advance, and have an alternative delivery path (a different platform or a direct HLS link) ready. This is the single most common reason a perfectly good stream "doesn't work for half the audience".
Geo-restrictions and privacy
Some content must not be visible outside certain territories or to the public. Platforms offer geo-restriction, password protection, SSO integration and tokenised links; unlisted YouTube links offer almost nothing. Decide the access model with the client and check the platform actually enforces it.
Redundant delivery
For events that matter we stream to two destinations at once — the primary platform and a standby (typically an unlisted stream on a second service). If the platform fails, the standby link goes to attendees by email within minutes. It costs almost nothing and has saved shows.