From Non-ICP Visitors to ICP Contacts

What it does

There's a category of RB2B row that gets discarded too quickly: the right company, the wrong person. A developer reading your documentation, an analyst comparing vendors, someone three levels below the budget holder.

Deleting those rows throws away the signal with the contact. Somebody at an account you want spent real time on your site, and that's worth acting on even though the person who did it isn't your buyer. This skill makes that explicit — it isolates exactly those visits and finds the ICP-matching people at the same companies.

When to use it

Run it when your identified traffic is good at the account level and disappointing at the person level. That pattern is common for technical products, anything with public documentation, and any product where evaluation starts well below the person who signs.

It's the natural follow-on to reading visitor intent signals: the signal tells you the account is in-market, and this turns that into someone to contact.

How it works

1. Person rows are read from the export.
2. You define your ICP — role, seniority, department — and any company-level fit criteria.
3. Visitors who don't match the ICP but whose companies do are separated out.
4. Those companies are deduplicated, so five visits from one account produce one search.
5. Each is searched for ICP-matching contacts, who are then enriched with email and optionally mobile.

Reading the output

The table pairs each visitor with the contact it produced: visitor name and title, company, then the contact's name, title, business email with risk grade, and mobile.

Keep both halves. The visitor is your reason for reaching out and the contact is who you reach out to, and an opener that references the actual signal — someone at their company has been evaluating this — lands very differently from one that arrives with no context at all.

One pattern is worth watching for: several different non-ICP visitors from the same account. One curious engineer is a data point. Four people from the same company reading the same pages over a fortnight is a buying group forming, and that account belongs at the top of the list regardless of what the individual titles look like.

Before you run it

You need a MoltSets account and the skill installed in Claude, plus an RB2B export that still has its LinkedIn URL column — that column is what every lookup keys off, so exports stripped down to names and emails won't work.

Run it on a small slice first. Twenty or thirty rows is enough to see the hit rate and the risk grades you're getting for your particular audience, and to decide which grades are worth keeping before you point it at a month of traffic.