Reading reply data.

What is signal. What is noise. What to chase.

Reply rate by angle is the headline. Look at the seven-day window, not the lifetime number. Lifetime smooths over the changes you actually want to see.

Positive replies

The number that matters is positive replies divided by sends, per angle. A 3% positive rate on a 14% gross reply rate is the same product as a 7% positive rate on a 9% gross reply rate. The first one is wasting your reviewers.

Negative replies

Negative replies are signal too. A wave of "wrong person" responses means the contact-resolution at enrich_contact() is drifting. A wave of "we already use X" means a competitor angle is missing from the pack.

When to retire an angle

The cleanest rule we have found: if positive replies under a given angle drop below the seven-day rolling p25 of that angle's history for two weeks running, retire it. The retire-an-angle playbook has the workflow.