7 Trade voice recording storage holds one of the most scrutinized data sets in a bank: the recorded conversations of traders, sales teams and advisers. Regulators require these calls to be captured, kept unaltered for years and replayed on demand. Compliance teams increasingly transcribe and analyze them to detect market abuse and misconduct. The storage underneath has to absorb continuous recording from turrets, desk phones, mobiles and softphones, keep every recording immutable for its retention period and make any call findable and playable within minutes, even years later. This article explains how trade voice recording works, how to estimate storage, what regulatory requirements imply for storage and how surveillance and analytics change the picture. For the broader context, see our hub on financial services archive storage. How trade voice recording works Trading floors use dealer boards, often called turrets, alongside desk phones, mobile phones and softphones in collaboration tools. Recording platforms capture calls from these sources, typically by tapping voice traffic through session border controllers or integrating with the telephony platform. Each recording is stored with metadata: participants, line or extension, timestamps, duration, desk and trader identity. Modern environments record: Turret lines, including always-on open lines and hoot-and-holler channels. Desk and mobile phones used for in-scope business. Softphone and collaboration platform calls. Video calls, increasingly, where used for client communication. The recording platform writes recordings to storage, often through an archiving layer that manages retention, holds and access. Estimating storage for voice Voice recordings are small individually but numerous and continuous. Storage depends on codec, channel count, recorded hours and retention. Codec bitrates Common telephony codecs differ widely in bitrate. Uncompressed G.711 runs at 64 kilobits per second, about 29 MB per hour of mono audio. Compressed codecs such as G.729 run at 8 kilobits per second, about 3.6 MB per hour. Many recording platforms store recordings in compressed formats, and stereo or dual-channel recording, which keeps each party on a separate channel for better transcription, doubles the data. A worked illustration Consider a trading floor with 2,000 recorded lines and devices, each averaging four hours of recorded audio per trading day, over 250 trading days a year, stored in a compressed format at about 4 MB per hour per channel, recorded in dual channel. Daily recorded audio: 2,000 x 4 hours = 8,000 hours. Daily storage: 8,000 hours x 8 MB = about 64 GB. Annual storage: 64 GB x 250 days = about 16 TB. Seven years of retention: about 112 TB, before protection overhead, copies, transcripts and indexes. Open lines recorded continuously, higher-bitrate codecs or video recording can raise this many times over. Larger institutions with many sites and channels reach petabyte scale, especially when video is included. Transcripts and metadata Transcription adds text files, timestamps and confidence data. Text is small compared with audio, but indexes for search and analytics can be substantial. Include them in capacity planning. Regulatory requirements that shape storage Retention: MiFID II and UK FCA rules generally require five years, extendable to seven on request. US rules for broker-dealers and swap dealers set their own periods. Other regions add local rules. Integrity: recordings must be protected from alteration or deletion. WORM storage or equivalent controls are commonly used. Completeness: gaps in recording can lead to regulatory findings, so storage must never be the reason a call was not captured. Retrieval: firms must be able to find and replay specific calls quickly for clients, investigations and regulators. Data location: some regulators expect recordings to stay in-country. Storage capabilities to look for Immutability Compliance-mode object lock ensures recordings cannot be altered or deleted before retention expires, even by administrators. High object counts Each call is a separate file, and some platforms split long open-line recordings into segments. Large estates generate billions of recordings over a retention period. Storage must handle very large object counts and metadata efficiently. Continuous write availability Recording never stops during trading hours. Storage must accept writes during drive failures, node failures and maintenance. Recording platforms typically buffer locally, but storage outages long enough to exhaust buffers risk gaps. Fast retrieval Searching happens in the archiving or surveillance application, but replay depends on storage delivering the recording promptly. Online object storage provides retrieval in seconds, compared with hours for offline media. Legal holds and retention management Holds for investigations must override normal expiry, and retention must be applied per jurisdiction. Regional deployment Global firms often run recording and storage regionally to meet residency rules and keep latency low, with consistent design and operations. Resilience Recordings should be protected within a site and replicated to a second site. Test recovery to ensure recordings remain replayable after a site failure. Surveillance and analytics Voice surveillance has changed from sampling a small fraction of calls to analyzing nearly all of them. Speech-to-text and language analytics now flag risky conversations for review. That has implications for storage: Read throughput: analytics engines read large volumes of audio for transcription, sometimes reprocessing historical recordings when models improve. Derived data: transcripts, embeddings and alert data add to capacity. Access control: analytics platforms need read access without the ability to alter or delete recordings. Data location: analytics often must run in the same region as recordings. Keeping recordings on scalable online storage, rather than offline archives, makes this practical. Mobile and remote working Since trading and advisory staff began working more often from home and on the move, mobile and softphone recording has grown sharply. Mobile capture may route calls through a carrier-level recording service or an app on approved devices, with recordings delivered to the firm’s archive. These recordings arrive from many sources with different formats and metadata quality, so the archive must normalize them and store them under the same retention and immutability rules as turret recordings. Firms should also monitor for capture failures, such as devices that have not delivered recordings for a period, since missing mobile records are a frequent finding in regulatory reviews. Operating the voice archive Day-to-day operations focus on completeness and availability. Useful practices include reconciling expected recordings against delivered recordings by line and day, alerting on recording gaps, monitoring storage capacity and write latency during trading hours, scheduling maintenance outside market hours and periodically replaying random recordings from each period to confirm they remain playable. Reports showing capture completeness and retention compliance help compliance teams respond to regulators quickly. Migration from legacy recorders Many firms still have recordings on older recording platforms with proprietary storage. Migrating to a modern archive must preserve audio quality, metadata, retention dates, holds and evidence of integrity, typically by hashing before and after migration. Some firms keep legacy recordings in place until retention expires and send new recordings to the new platform, avoiding large migrations. Security of recordings Recordings can contain market-sensitive information and personal data, so encrypt them at rest and in transit, restrict replay rights to authorized compliance and legal staff and log every replay and export. Checklist: trade voice recording storage Inventory recorded sources: turrets, desk phones, mobiles, softphones and video. Estimate capacity from codecs, channels, recorded hours and retention. Include transcripts, indexes and analytics data. Store recordings with compliance-mode immutability. Ensure continuous write availability through failures and maintenance. Deliver fast retrieval and replay from online storage. Apply retention per jurisdiction and support legal holds. Deploy regionally for residency and latency. Provide read access for surveillance without write or delete rights. Plan migrations from legacy recorders with integrity evidence. Putting it together Trade voice recording storage must capture every call, keep it unaltered for years and replay it on demand, while supporting surveillance that now analyzes nearly every conversation. Voice volumes are modest per call but large in total, and video and analytics push them higher. Compliance-grade object lock, high object-count scalability, continuous write availability, fast retrieval and regional deployment make object storage a strong foundation for modern voice archives. Frequently asked questions How much storage does trade voice recording need? It depends on codec, channels, recorded hours and retention. A floor with 2,000 lines recording four hours a day in compressed dual-channel audio might need around 16 TB a year before protection overhead. How long must trader calls be kept? Under MiFID II and UK rules, generally five years and up to seven on request. US and other regional rules set their own periods. Do voice recordings need WORM storage? Recordings must be protected from alteration and deletion. WORM storage with compliance-mode object lock is a common way to meet that requirement. How does voice surveillance affect storage? It increases read throughput and adds transcripts, indexes and analytics data, favoring online, scalable storage. Should old recordings be migrated to new platforms? Sometimes. Many firms leave legacy recordings in place until retention expires and send new recordings to the new platform. Further reading Financial services archive storage MiFID II and Dodd-Frank records storage Insurance document retention storage SEC 17a-4 compliance
