What it does
The role agent for lifecycle and retention marketing. It works on what you paste — an ESP flow export, a drip definition, a cart table, a billing event log. It reaches no platform, holds no credential and has no clock. Four capabilities. read_flow_performance reports each step of a flow in SEND ORDER with the 95% interval its counts support, and flags a step only against a target YOU state. check_drip_structure finds the defects decidable from the definition alone — a flow with no exit, an exit set that omits the flow's own goal, two steps at the same hour, a step stranded past a hard time limit. read_cart_recovery reports recovery by segment and credits an effect to your email only where you supplied a holdout. read_subscription_transitions counts plan changes and reports the mean MRR step already observed for each trigger, with its standard error beside it. What it will not do is the point. There is no built-in click or open-rate threshold. A 2% click rate is poor for a receipt and excellent for a cold reactivation, and nothing in an export tells them apart — so a verdict against a fixed constant is a property of the constant, not of your flow. It will not rank steps against the flow's own median either. That is the obvious replacement and it is worse: half of any set is below its own median by construction, so a perfectly healthy flow would report half its steps as problems, every time. What can be said honestly is whether two steps differ by more than the counts resolve. It reports open rate and never judges on it. Apple Mail Privacy Protection fetches the tracking pixel whether or not a human looked, so an open rate measures a mail client mixed with a person in a proportion nobody sending the campaign can observe. And it will not credit a recovery to your email without a holdout. Some proportion of abandoned carts return on their own, and every one of them lands in the numerator of "carts recovered after we sent the email". No credentials. Every capability reads data you paste, so there is no key to store, no host to reach and nothing to leak.
Reads lifecycle email against the counts, not against a benchmark nobody can source
Example prompts
- Here are my welcome flow's steps with sends, opens and clicks - which ones actually differ, and which can these volumes not tell apart?
- My day-3 email has a 1.2% click rate. Is that bad?
- Here is my abandoned-cart flow definition - can it keep emailing someone who already bought?