Sentiment analysis that never stores the message
SentiTrack scores the tone of every email, call and chat 1–10, then throws the body away. Your database keeps only the score, the participants, a timestamp, the direction and a subject hash — never content.
No-body storage means SentiTrack scores each message's tone 1–10 in memory, then discards the body. It persists only the score plus non-content metadata — participants, timestamp, direction and a subject hash — never the body, subject text or attachments. Zero message bodies are stored, by design.
Tracking tone usually means hoarding content
Most sentiment tools keep the text they analyze, turning your insight project into a data-breach liability and a subpoena target. Storing message bodies tied to identifiable employees — especially alongside demographic attributes — is exactly the personal-data exposure a privacy-conscious B2B buyer wants to avoid. The more you store, the more you have to defend.
Keep the insight, drop the content
SentiTrack inverts the usual trade-off. Bodies are sent to your chosen AI provider only to compute a single 1–10 score and a stable topic, then they're discarded in memory. What lands in your database is just the number and non-content metadata — enough to power every graph, timeline and report, with nothing sensitive to leak.
- Bodies scored in memory, then discarded — never written to disk
- Only score, participants, timestamp, direction and subject hash stored
- Subjects kept as a one-way hash — never plaintext
- Same posture across email, voice, chat and the /ingest API
How no-body storage works
Each message's plaintext is sent to your chosen AI provider — bundled SentiAI or your own Anthropic, OpenAI or Gemini key — to compute a 1–10 sentiment score and a stable topic.
As soon as the score returns, the body, subject text and attachments are dropped from memory. Nothing content-bearing is ever written to your database.
Your database stores the 1–10 score, the participants, a timestamp, the direction and a subject hash — the non-content metadata that powers every dashboard, with nothing to leak.
Because bodies are never persisted, there's no message archive to breach, subpoena or mishandle — only aggregate 1–10 scores.
Storing only a score and metadata shrinks your data-protection surface versus tools that retain full message text.
Bring-your-own Anthropic, OpenAI or Gemini keys are encrypted at rest per organization with Fernet — never shared across tenants.
A configurable retention window (365 days by default) ages out old scores automatically, on your schedule.
A hard-delete endpoint removes an organization's scored records outright when you need a clean slate or honour a request.
Run SentiTrack Edge in your own network so message content never leaves it — only finished scores are pushed to your account.
We score the body, then throw it away
Storing sentiment alongside identifiable employees and demographic attributes can be special-category data under GDPR Article 9. SentiTrack minimizes that exposure by never persisting content: your database holds only a 1–10 score and non-content metadata, with subjects reduced to a one-way hash.
- Zero message bodies retained — scored in memory, then discarded
- Zero subjects stored in plaintext (hash only)
- Per-org Fernet-encrypted keys for bring-your-own-AI
- Hard-delete endpoint plus a configurable retention window (365-day default)
Privacy & no-body storage — FAQ
Does SentiTrack store email or message content?
No. SentiTrack sends a message's body to your chosen AI provider only to compute a 1–10 sentiment score, then discards it. Your database stores just the score, the participants, a timestamp, the direction and a subject hash — never the body, subject text or attachments.
What exactly does SentiTrack keep about each message?
SentiTrack stores only non-content metadata per message: the 1–10 sentiment score, the participants, a timestamp, the direction (internal vs external) and a one-way hash of the subject. The body, plaintext subject and attachments are scored in memory and then discarded, so nothing content-bearing is retained.
Is SentiTrack GDPR-compliant?
SentiTrack is built to minimize personal-data exposure by never persisting message content — only a 1–10 score and metadata. Storing sentiment alongside demographic attributes can be GDPR Article 9 special-category data; the lawful basis, any DPIA and employee consent remain the customer's responsibility under the Terms of Service, since the customer decides what is monitored.
How long does SentiTrack retain data?
Each organization sets its own retention window, defaulting to 365 days, after which scored records age out automatically. You can also call a hard-delete endpoint to remove your organization's scored records outright at any time. Since only scores and metadata are ever stored, there is no message content to retain.
How are my AI provider keys protected?
If you bring your own Anthropic, OpenAI or Google Gemini key, it is encrypted at rest per organization using Fernet symmetric encryption and never shared between tenants. If you use the bundled SentiAI engine, no key is required at all. You can switch provider, model or key in Settings.
Can I keep message content entirely inside my own network?
Yes. SentiTrack Edge is a Docker appliance you run inside your own network that ingests and scores locally, pushing only finished 1–10 scores and non-content metadata to your hosted account every 12 hours. Content never leaves your network, at roughly 93.5% of the full hosted model's accuracy (approximate, not certified).
Get the tone signal without the content liability
Explore the live demo with sample data — no signup — or connect a channel and watch scores fill in while bodies are discarded. Month-to-month, cancel anytime.