What call recording actually is (and what it isn’t)

Call recording captures the audio. It is not the same as your call logs, and it is not the same as your compliance policy, though the three are often confused.

Voice logging and call recording are the same thing: a system that captures the spoken audio of a call and saves it so you can play it back later. The recording is stored as an audio file (typically MP3, WAV or Opus) and tagged with metadata such as the numbers involved, the time, the duration and the agent or extension. That playable audio is the point of difference from everything else your phone system produces.

Two neighbouring capabilities get muddled with it, and keeping them apart saves you money and licensing headaches:

• Call accounting / CDR data is the metadata only: who called whom, when, for how long and at what cost. It answers “how much are we spending and where”, not “what was said”. That is the job of a telephone management system, and you can have it without recording a single second of audio.
• Recording compliance is the legal wrapper: consent, lawful basis, security and retention. The recorder captures the audio; your policy decides whether you are allowed to, and for how long you keep it. We cover that in depth in the POPIA-compliant call recording checklist.

How a call gets recorded: three capture paths

There are only three places a voice stream can be tapped: in the network, on the platform, or at the endpoint. Where the copy is made decides how much you can record and how well it scales.

A modern VoIP call is just a stream of audio packets between two devices. To record it, a recorder has to sit somewhere it can see that stream and copy it, without breaking the live call. The three architectures below cover almost every deployment in South Africa.

1. Network capture (SIPREC)

A session border controller or SIP proxy in the call path forks a copy of the media to a recording server. This is standardised by the open SIPREC protocol (IETF RFC 7866, with the metadata model in RFC 7865), so the recorder does not have to be from the same vendor as the phone system. It scales cleanly, survives platform changes and is the enterprise default.

2. Cloud PBX built-in recording

Most hosted phone systems now record natively: you toggle it per extension or per queue in the portal, and the audio lands in the provider’s storage. It is the fastest to switch on and needs no extra hardware, which is why it suits most South African SMEs. The trade-off is that your recordings live inside one platform, so check export and portability before you commit.

3. Endpoint or gateway capture

The oldest pattern records at the device: on the handset or softphone, or via a passive tap on a mirror (SPAN) port on the LAN. It still has a place for on-premise PBXs and analogue lines, but it is fiddly to manage at scale and is fading as platforms move to the cloud.

Why South African businesses record calls

Recording earns its keep in five concrete jobs. If your reason isn’t on this list, you probably don’t need it on every line.

Beyond those, recordings feed fraud and security investigations and, increasingly, AI analysis: sentiment, keyword spotting and automated QA that reads every call instead of the 2% a supervisor has time to sample. The caution is the same for all of them: a recording you never review, secure or delete is cost and risk with no return.

Storage, retention and keeping it legal

A recording has a lifecycle. Get the last two stages right, retention and disposal, and most of the compliance risk takes care of itself.

Under POPIA a recording that contains someone’s voice or personal details is personal information, so you need a lawful basis to make it, you must secure it, and you must not keep it longer than you need it. Under RICA, recording a call is an interception, which is lawful when you are a party to the call or record it in the ordinary course of business. The full checklist lives in our POPIA-compliant call recording guide; the lifecycle below is the operational shape of it.

How much storage will it take?

Sizing is simple maths. A standard G.711 voice call is 64 kbps, so one uncompressed hour is roughly 28 to 30 MB; stored as MP3 or Opus that drops to a few MB per hour. A ten-seat team on the phone four hours a day generates on the order of a few gigabytes a month uncompressed, which is trivial for a cloud platform but worth checking against any per-minute storage fee.

How to choose a call-recording solution

Work from the requirement backwards. The right setup is the smallest one that meets your legal duty and lets you find a call when you need it.

Keep reading