Twitch on the panels: here, viewers are the product

On Twitch, the live broadcast is not one format among many, it is the product. That shapes the stock: across all indexed panels there are 3,940 orderable entries in 8 service tiers from 25 catalogues, and the viewers during a live stream make up the largest part of them.

That is why the goods are simply called viewers here, while the same goods on Instagram or YouTube have to be called "live viewers" so they are not confused with the view count. Conversely, views on recorded videos on Twitch carry their own name, video views. Both labels are the same distinction seen from two sides.

Viewers are sold by duration, in tiers from minutes up to many hours, and this tiering is strikingly similar across providers. Anyone comparing should therefore compare within one tier: same duration, same quantity, then the price.

One figure still belongs to the picture, because it comes from the fields and not from the titles: the number of offers in this network where the panel reports a refill through its own API is 41. This refers to the self-triggerable variant; what the offer name states is a promise, not a field. For the viewers of a live stream, however, a refill is moot: once the broadcast is over, there is nowhere for it to go.

The lowest priced base rate in this network is 0.009 € per thousand units, and the middle of the range is at 14.17 €. The lower figure is currently set by SocNet — that is a property of a single catalogue, not a market situation, and at the next pull it may sit with a different provider. Both are purchase prices at wholesale level, not what a German-language sales page charges for them.

Viewers are the goods of the live stream: they must be present during the broadcast, and a refill afterwards has no place left to go.

Followers stay on the channel, regardless of whether a broadcast is currently running. Video views count plays of recorded content and are explicitly not the same number as the viewers above. Chatters are chat messages — the only goods here that visitors read rather than count, and correspondingly conspicuous when they do not fit what was said.

Which providers stand behind these lines is set out in the panel comparison.

Between two pulls the catalogue does not stand still. Since 31/10/2026, 125 entries in this scope have a different base price than before, of which 0 upward and 125 downward; 0 rows have been newly added and 0 have disappeared from the catalogues. The date given is that of the oldest of these earlier pulls, so the line does not claim a freshness its slowest component does not have.

Comparison only covers the panels that have a second pull at all; their number is 25 and is not the number of indexed panels, because a newly added provider has no history yet. As long as this second pull is missing, nothing appears here — a zero would mean "nothing has moved", and that is a measurement, not a gap.

Two details appear in no panel API and are therefore maintained by hand from the panels' own statements: whether a provider issues a proper invoice and which payment methods it accepts. The number of providers with an invoice in this scope is 2. Since 30/08/2026, none of these details here has gone unchecked — the oldest check is shown, not the most recent, because a date based on the last check would hide the providers that nobody has looked at for longest. The number of providers with no check at all is currently 0.

Many of these rows resemble one another down to the quantity limits, and there is a reason for that: panels buy from the same upstream suppliers. Where entries from different providers within a tier exceed our similarity threshold, we list them as a group and mark the more expensive rows with a note that they probably sell the same source — not all of them, because where the price gap can only be a counting error, the note is left out. The number of groups touching this section is 820, the number of marked rows is 1,187.

"Probably" carries this sentence. No panel writes into its API who it buys from; we see two rows that resemble each other, and that is a hint, not proof. The groups also do not divide the stock — what is counted is what touches this section. The measurement is described in detail in the panel comparison.