How do you build a signal-based outbound list?

A signal-based outbound list is built in two steps. First you decide which accounts fit your business, using your own closed-won deals. Then you watch only those accounts for events that make a purchase likelier this week, such as a relevant hire, a funding round or a job change, and you contact the ones where an event has fired.

The rest of this post sets out which signals earn a place, which do not, and how the list is put together.

Which signals are worth watching?

A signal is worth watching when it is public, recent, tied to what you sell, and checkable by a person. Four kinds meet that test for a company that sells to other businesses: hiring for a role your product serves, a funding round, a change of job by a buyer or champion, and activity on your own site or content.

Each one tells you something different, so each gets its own play.

• A relevant hire. A company posting for a role your product supports has usually already decided it has the problem. If you sell to revenue teams, a posting for a revenue operations lead is a stronger reason to reach out than a posting for a designer. Our hiring signal feed shows how we rank postings by title.
• A funding round. New money is a budget window. In our library, the funding recipe waits 30 to 45 days after the announcement before the first touch, so the message arrives after the press release noise has settled.
• A job change. A former user or champion who joins a new company is a warm introduction you did not have to earn twice.
• Your own data. A visit to your pricing page, a reply to an old campaign, or a lost deal that has since changed. These are the signals nobody else can sell you.

Which signals are noise?

A signal is noise when it is true of nearly every company, cannot be checked, or has no link to your offer. A "company is growing" flag, a vague intent score you cannot trace and a posting for a role unrelated to your product all produce lists that look targeted and are not.

Three tests catch most of it.

1. Would this be true of almost anyone in the market? If so, it does not separate one account from another.
2. Can a person open the source and see it? A posting, a press release and a profile can be checked. A score from a black box cannot.
3. Does it change what you would say? If the event would not alter the first sentence of your message, it is not a reason to write.

Stale events fail the second test quietly. A posting that has been up for two months says less about this week than one posted yesterday, so age belongs in the rules, not in somebody's memory.

Where do the signals come from?

Signals come from three places: public data such as job boards, funding news and job changes; custom scrapes for events nobody sells; and your own CRM and product data. Teams usually start with public data because it is available on day one, then add their own events as they learn what predicts a deal.

Custom scrapes are worth building when your buyer leaves a public trail that no vendor tracks, such as a regulatory filing, a store opening or a change on a careers page. We build these with Claude Code and write each result to the same table as every other signal, so the list reads as one feed.

Your own data is the source nobody else can sell you, and teams often reach for it last. Your closed-won and closed-lost deals show what a good customer looked like at the moment they bought, and that pattern tells you which public signals to trust.

How do you combine a signal with fit?

Treat fit and timing as separate questions. Fit asks whether the company is one you want as a customer, and it is answered once, from your own deals. Timing asks whether something has happened lately, and it is answered every week. A signal moves an account up inside its tier. It never changes the tier.

This keeps the list honest. Without the split, a loud signal at a poor-fit company floats to the top of the list and a rep spends a morning on a deal that was never going to close.

In practice the map comes first. The ICP map walks through turning a raw list into tiers, with the reason for every exclusion written down. Then each signal is scored on how recent it is, and the freshest events at the best-fit accounts are worked first. When a signal is missing a fact, such as a posting with no date, it waits for a person to look at it and is never treated as a pass.

How many signals do you need, and how do you keep the list clean?

You need as many as one person or one sequence can act on in a week, and no more. A list is a work queue, so its right length is the number of accounts your team can contact well. Keep it short, rank it by timing, and record why every account entered or left.

Three habits keep it clean.

• One event per company, per signal type, per day. Twelve postings for one role are one signal, not twelve.
• Write the rules in one file. The titles that count, the ones that never count, and how fast a signal ages. When someone asks why a company was on the list, the answer is in the file.
• Keep the exits. Every company that drops out leaves a reason, such as too small, not B2B or already a customer. Each reason becomes a filter for the next pull.

This is the system we build in GTM engineering: the map, the signals, the enrichment, the plays, and the record in your CRM, handed to your team at the end.

What should you try first?

Start with one signal and a list short enough to read in full. Take your last ten closed-won deals and write down what was true of each company in the month before it bought. Pick the one event that appears most often and that you can actually observe, and watch it for two weeks.

Contact only the accounts where it fires, and write down what you said and why. If the replies are better than your cold list, add a second signal. If not, the signal was noise, and you have learned that cheaply.

You will also find that the exercise changes what you ask of any vendor. The question stops being how many contacts a database holds and becomes which of them have a reason to talk to you this week.