Pre-send message testing
Test before
you send,
not after.
Put message variants in front of real readers, measure what they do and what they say, and make an evidence-backed release decision.
Interactive sample
Variant comparison
Select a variant to inspect its evidence
Sample study data. Variant A selected — Control. Subject line: “Your account is ready — the three things to do first”. Trust 4.7, Clarity 4.7, Action intent 4.3, CTA clicks 6, based on 6 responses. Release gate: GO — All conditions met. Recommendation: Send Variant A — A outperforms B on trust, clarity and action intent, with 3× the CTA clicks across the same number of responses.
How it works
Build two to five variants — subject, preview text, body and call to action — in one test.
One tokenized link per variant. Reviewers open it and read. No account, no panel, no install.
Five behavioral events as they read, then four questions once they finish.
The release gate returns GO, CONDITIONAL, WAIT or STOP — and shows the evidence it used.
Methodology
How SigSent turns responses into a decision.
Behavioral signals
See how people actually engage, not only what they say afterward.
Opens, scroll depth at 25 / 50 / 75 / 100, CTA clicks, completion and abandon — recorded per session and attached to one variant.
See in sampleDirect feedback
Ask the right questions while the message is still fresh.
Trust, clarity and action intent on a five-point scale, plus a recall question and an optional comment. One response per session, no double-counting.
See in sampleRelease gates
Define your bar. SigSent tells you plainly whether it was met.
Three responses before an early leader is named, five before the gate returns anything but WAIT. When the gap is under threshold it says too close to call, not “winner”.
See in sampleEngineering
AI grounded in human evidence.
AI drafts and refines. Reviewers decide. Every rewrite is tested the same way the original was.
Thresholds are fixed rules, not model judgment. The same evidence always returns the same verdict.
Suggestions cite the sessions, scores and quotes behind them — your reviewers, not a model’s taste.
Security
What we protect, and what we do not collect
Reviewers are never asked for a name or email; we do not record IP location. There are no advertising or cross-site tracking cookies.
Credentials and secrets are encrypted at rest with AES-256-GCM.
Review links expire on the date set when they were created, up to 30 days. Named processors are listed in Privacy.
Works with the tools you already use
Webhooks
Mailchimp
Send the version people actually respond to.
Test with real readers. Raise your signal. Reduce your risk.