WebRTC makes a two-person video call surprisingly easy to prototype. Browsers handle capture, encoding and peer connection, and a working demo comes together quickly. That demo creates an expectation about effort that the rest of the work does not meet.
The first surprise is connectivity. Direct peer connections fail behind many corporate networks and some mobile carriers, and the fallback requires relay servers that carry the media. Running those servers is an ongoing operational and bandwidth cost that the prototype, working on an office network, never revealed.
The second is group calls. Two participants is a peer connection; eight participants is a media server routing streams, deciding which to forward, and adapting to each participant's bandwidth. That is a different class of system, and it is the point at which most teams stop building and start buying.
Managed platforms — Twilio, Agora, Daily, LiveKit among others — handle the infrastructure and charge per participant-minute. The pricing looks modest until you model actual usage, and it is worth doing that arithmetic before committing, particularly for products where calls are long or frequent.
Recording brings its own weight. Storage grows quickly, and a recorded call is personal data that needs a retention period, an access model and, under the DPDP framework, a lawful basis and usually a clear notice to participants. Products that added recording as a checkbox often discover the compliance surface afterwards.
Before any of this, ask whether the call needs to happen inside your product at all. For many workflows, generating a link to an existing meeting platform delivers the outcome with none of the engineering. In-product calling is worth building when the call is tied to product context — a support session showing the user's own data, a consultation with records alongside — and hard to justify otherwise.
