Radar, What worked
What worked answers one question: which arguments moved which audiences, through which vehicles. It measures the message, not the email.
The strip
Three lines at the top. The biggest gain: a claim that ran ahead for one audience. The biggest loss: a claim that ran behind for one. The least tested audience: the target with the most known contacts and the fewest claims tested. Each line names the next move. A line shows only when the interval clears the audience's own rate.
Three lenses
By audience: one line per target. Tap it for the top claim, the lowest, the appeal that worked, the ladder, and the trait wheel. The solid shape is observed from results. The dashed shape is inferred from the card. By argument: one line per claim, then per appeal, with Next moves. By vehicle: one line per pool with its messages, then the articles contacts reached from a send.
Tag first
Open Blocks, open an entry, look under Claims this entry states. Confirm a claim, or choose Not this. A tag holds for that exact wording. Edit the entry and new sends need the tag confirmed again.
Reading a claim line
One line per claim: the claim, the share of contacts who clicked, the contacts reached. Tap for the range, the moves, unsubscribes and replies. A rate shows from 30 contacts. Ahead and Behind mean the range clears every other claim. Otherwise Cannot tell.
The top line says how many sends in the window carried a confirmed claim. Under 80%, read the results as a hint and tag more entries.
Next moves
Suggest picks up to three moves, in code: lead with a claim, rework one, or answer an objection. The AI words them. A sentence with a number not in the facts is refused. Each move is a 30-day prediction, graded in code: Held, Missed, Cannot tell, Too early. The hit rate shows after 10 graded.
By audience
Each send is matched to one target (ICP) in code. The match uses the contact's title, industry, company size and country against the target cards. A contact the rules cannot place is read once by a model, per set of facts. Pick an audience at the top to see the claims for that target alone. The picker lists the targets with results in the window. Nothing is scored or stored on a person. The ICP is a group.
Appeals
An entry can carry one appeal: Functional, Emotional, Signal or Proof. Set it on the entry under Blocks. Appeals shows one line per appeal with the same numbers as a claim.
Articles
A hub article can state claims too. Tag them on the article's Claims tab. Articles shows each article contacts reached from a send: how many, and how many converted or moved after. The article's claims get that credit.
Objections
Objections lists up to five themes from the objection replies of the last 30 days. A model groups the replies; the count is taken in code. A theme is a phrase, never a quote. Themes appear only after 50 reply labels are checked and agree well enough. By audience shows each target's own themes.
The ladder
Each claim also counts what contacts did after the send, from HubSpot.
- Responded: a reply date or sales activity.
- Converted: a form, a gate, a meeting.
- Advanced: the lead status changed.
- Pipeline: a deal created or won.
Read nightly. A step HubSpot has not granted reads zero; the Integrations card says which permission it needs.
Replies
HubSpot logs a reply on the contact when your email's Reply-to is a connected inbox. Turn on Replies under Settings, Integrations. It needs one HubSpot permission; the card says when to reconnect. mxAURA then reads new replies every hour and ties each to the send it answered. A model labels them. The owner or an admin checks labels on the same page. Reply mixes stay unvalidated until 50 are checked and agree well enough. Nothing is forwarded; nothing leaves HubSpot but the reply's own words, kept 30 days.
After the send, not caused by it. Results build nightly.