Recap
- Three things make a signal predictive: how recent the trigger is, whether the person owns the problem, and whether the trigger actually relates to what you sell.
- Most company-level intent scores are weak on their own. They tell you an account is warm, not who to email or why now.
- Opens are noise. Apple Mail Privacy Protection inflates them. Weight replies, meetings, and specific clicks instead.
- Signals improve odds, they are not magic. And silence in a feed usually means the source does not cover your market, not that demand is gone.
The signals that predict a reply are recent, aimed at the right role, and tied to your product. A director of support hired three weeks ago is a strong reason to email a support tool. A company that 'showed intent' for your category last quarter, with no name and no event attached, is barely a reason at all. The difference is specificity. Good signals point at a person and a moment. Weak ones point at a logo and a guess.
Which signals actually correlate with replies?
The ones that combine recency, role fit, and product relevance. A trigger that happened in the last two weeks, landing on a person who owns the problem you solve, where the trigger and your product are clearly related, is the strongest setup there is. Miss any one of those three and the odds drop fast.
Think of it as three filters stacked on top of each other. Recency answers why now. Role fit answers why this person. Relevance answers why us. A new funding round (recent) reaching the VP of Engineering (role) about a hiring or infrastructure product (relevant) checks all three. The same round reaching a junior marketer about an unrelated tool checks one.
What separates signal from noise?
Noise is anything that is stale, aggregate, or disconnected from your offer. A company-level intent score with no contact and no event is the classic example. It feels like data, but you cannot write a first line from it. If you cannot turn a signal into one honest sentence about why you are reaching out today, it is noise.
Here is how common signals tend to sort out. The point is not that the weak ones are useless, it is that they belong in a ranking layer, not as the reason for the email.
| Signal | Why now | Points at a person | Strength |
|---|---|---|---|
| New hire in the owning role | Days to weeks | Yes | Strong |
| Open job req for the role you serve | Weeks | Mostly | Strong |
| Funding round (relevant to product) | Weeks | No, you infer it | Medium |
| Tech stack change or new tool adopted | Weeks | Sometimes | Medium |
| Visited your pricing page | Hours to days | Sometimes | Strong when known |
| Company-level intent score | Vague | No | Weak alone |
| Generic firmographic match | None | No | Noise on its own |
How recent does a trigger have to be?
Fresher is almost always better, and the useful window is shorter than most teams assume. A trigger loses most of its predictive power within a month, and a lot of it within two weeks. The person who just inherited a problem is open to ideas. The same person three months in has already picked a tool or moved on.
There is a floor too. A signal that is hours old, like a pricing-page visit, can be too hot to play casually. Reaching out within minutes can feel like surveillance. The sweet spot for most cold motions is acting while the trigger is still the current reality for the buyer, which usually means days, not minutes and not months.
How should I weight these signals?
Score on the three axes and multiply, do not add. Recency, role fit, and relevance should each gate the others, because a perfect score on two and a zero on the third should kill the send. A recent, perfectly relevant trigger aimed at someone who does not own the decision is still a bad email.
- Recency. Decay it. Full weight in the first two weeks, half by a month, near zero past a quarter.
- Role fit. Treat it as close to binary. Either this person plausibly owns or influences the problem, or they do not.
- Relevance. Ask whether you can name the connection between the trigger and your product in one sentence. If you cannot, the relevance score is low no matter how exciting the trigger looks.
- Stacking. Two independent weak signals pointing the same way (a job req plus a tech change) can beat one medium signal. Independence is what matters, not count.
The output of all this is not a yes or no. It is an ordering. You contact the strongest setups first and let the weak-on-their-own signals decide ties.
What signals are mostly noise?
Anything that describes a category instead of a moment. Pure firmographics (industry, size, region) tell you who could buy, never who is ready. Aggregate intent topics without a person attached are the same trap dressed up as data. And opens, despite how good they look on a dashboard, are close to meaningless now.
Opens deserve a special callout. Apple Mail Privacy Protection pre-fetches images on the recipient's behalf, so an open often registers whether or not a human ever read the message. If you are optimizing copy or timing off open rate, you are tuning against a number that no longer means what it used to. Judge interest by replies, booked meetings, and clicks on something specific.
What signals will not do for you
Signals improve your odds. They do not create demand, and they do not excuse a weak message. A flawless trigger sent to the right person still fails if the email ignores the trigger entirely and reads like every other template. The signal earns you the right line. You still have to write it.
The honest caveat is about absence. Most signal sources lean heavily toward tech, SaaS, and consumer markets, because that is where public footprints are richest. If you sell into a niche, regulated, or largely offline vertical and a feed shows nothing, the safe read is that the source does not cover your buyers, not that the buyers went away. Silence is a gap in coverage far more often than it is a verdict on demand. This is exactly the kind of weighting and decay that an autonomous operator like LaunchSurface bakes into who it contacts and when, so a stale score never outranks a fresh, role-relevant trigger. Treat signals as a way to be early and relevant, never as proof that the market is or is not there.