A signal is not useful because it is observable, it is useful because it changes the probability of a reply. Three properties decide: freshness, scarcity, and distance to the problem you solve. A funding round is fresh and public, so everyone acts on it at once and its value collapses. Rank your signals on those three axes before buying any tool.
Teams that move to signal-based selling usually start by making a list of everything they can detect. Funding rounds, hires, tool changes, page visits, LinkedIn posts, job ads, leadership changes. The list gets long, the tooling gets expensive, and the reply rate barely moves.
The problem is that being observable and being predictive are two different things. Most of what a prospecting tool can detect tells you something true about a company and nothing about whether it will answer you this month.
Three properties decide the value of a signal
We rank every signal on the same three axes before deciding whether it earns a place in the motion.
Freshness. How long has the event been visible? A signal loses most of its value within days. The buying window it points to is open now, and it closes while your list is being cleaned.
Scarcity. How many other vendors see the same thing at the same time? This is the axis teams ignore, and it is the one that decides whether your message stands out or lands in a pile of eleven identical ones.
Distance to your problem. Does the event touch what you actually solve, or is it just corporate news? A company opening an office is interesting for a facilities vendor and irrelevant for a compliance tool.
A signal that scores high on all three is rare and worth building a sequence around. A signal that scores low on all three is a distraction with a data subscription attached.
Why the most popular signals convert the worst
Funding rounds are the clearest example. They are fresh, public, dated to the day, and every prospecting platform indexes them within hours. A company that announces a round receives a wave of near-identical messages that same week, all opening with a congratulation. The signal is genuine, its scarcity is zero, and scarcity is what buys the reply.
The same holds for the obvious technology signals. If a tag is detectable from the outside, it is detectable by everyone. Knowing that a company installed a given analytics script puts you in a queue with every vendor whose positioning mentions that script.
This does not make public signals useless. It makes them poor as a trigger and useful as a qualifier: not a reason to write today, but a reason to prioritise one account over another when both already fit your ideal customer profile.
The signals that keep working
Three categories hold up because they are harder to see from the outside.
Movement inside the team that owns your problem. Not a leadership change in general, but a departure, an arrival or a reorganisation in the exact function that would buy you. A new head of the team that owns your problem rewrites their stack in their first hundred days, and almost nobody is watching that specific seat.
Behaviour on your own properties. A second visit to your pricing page, a documentation page read twice, a comparison page opened. It is the scarcest signal that exists, because you are the only one who can see it. It is also the one most teams leave unused because it sits in an analytics tool nobody has connected to the CRM, a gap we cover in building your revenue operations foundation.
Dated constraints. A regulatory deadline, a contract renewal, an end of support, a certification expiry. These are not events you detect, they are dates you compute. They are unglamorous and they convert, because the prospect has a problem with a calendar attached.
Rank before you buy
The order matters. Most teams buy a data source and then look for signals inside it, which is how you end up paying for a feed of funding rounds.
Score your candidate signals on the three axes first, on paper, with the team that will act on them. You will usually find that your two best signals are already in your own systems and cost nothing, and that the paid feed is worth buying for one specific trigger rather than for its catalogue. That framing changes the conversation with vendors, and it is the same reasoning we apply to channel choice in inbound, outbound, allbound: dropping the false debate.
Measure per signal, not per campaign
A signal-based motion that reports a single reply rate cannot be improved. You will know the machine works or does not, and never which part.
Tag the triggering signal on every sequence entry, and read the reply rate by signal. Thirty to fifty attempts per signal is enough to see the difference between a trigger that earns a conversation and one that produces polite silence.
Then apply the rule that keeps the motion honest: any signal that fails to beat your ordinary sequence gets dropped. Not improved, dropped. A trigger that needs three rewrites to reach baseline is telling you it was never close enough to your problem.
This measurement discipline is also what stops the motion from quietly decaying once the person who built it moves on, which is the failure mode we describe in a sales playbook that survives departures.
The workshop that produces your inventory
The ranking happens in one session, not in three weeks of reflection. We run it in ninety minutes, with the reps who prospect, the person who owns the CRM, and someone from product or support who knows what customers actually live through.
It runs in four parts.
Twenty minutes of listing. Everyone writes down the events that, in their experience, preceded a won deal. Nothing is filtered, nothing is debated yet. Reps surface things no tool catalogue would have produced, because they lived them.
Thirty minutes of scoring. Each signal gets three scores from one to three: freshness, scarcity, distance to the problem. Scoring is done openly, and disagreement is the most useful moment of the session. When one rep scores a signal three on scarcity and another scores it one, one of them knows something the other does not.
Twenty minutes of sorting. Keep the signals reaching seven out of nine, drop the rest. The rule is mechanical on purpose: it prevents the discussion turning on the favourite signal of the most insistent person in the room.
Twenty minutes of attachment. For each surviving signal, one sentence answers the question: what do we say to someone this just happened to, that we would not say otherwise? If nobody can find the sentence, the signal goes, whatever its scores. A trigger without its own message is not a trigger, it is a pretext.
You leave with three to five signals, their source, and the opening line attached to each. That is enough to start, and it is deliberately few.
The three most common failures
The signal is detected and nobody acts on it. This is failure number one, and it is organisational rather than technical. The signal lands in a notification no one is responsible for. Name an owner per signal, and set the delay within which it must be handled. Without a name and a deadline, a detected signal is a lost signal.
The message says nothing about the signal. A team sets up detection, then sends its usual sequence to the detected accounts. Reply rate does not move, and the conclusion drawn is that signals do not work. What does not work is changing the targeting without changing what you say. The message should be incomprehensible to someone the event did not happen to.
Volume collapses and the criteria get loosened. After a few weeks, well-scored signals produce only a handful of accounts per week, which is normal since they are scarce. The usual reaction is to lower the bar until volume returns, and the motion becomes a list again. If volume is short, add another properly scored signal rather than degrading the ones that work, or accept working fewer accounts with markedly more care.
Where to start
Take your last ten won deals and work backwards: what had happened at the customer in the thirty days before the first useful conversation? In most teams, two patterns come back, and neither appears in any tool catalogue. Those are your first two signals, and they cost nothing.
Then give yourself a quarter before judging. A signal-based motion produces fewer conversations than a list in its first weeks, and better ones. Teams that abandon it usually do so in week three, on a volume comparison that was never the point.
Where this sits in the method
Ranking signals is one step of a larger motion: detecting the event, deciding the window, writing the message that only makes sense because of that event. The full sequence is in signal-based selling: the complete method.
If you only take one thing from this article: stop asking whether a signal is true, and start asking who else can see it.
Frequently asked questions
- What is a buying signal?
- An observable event suggesting a company has just entered, or is about to enter, a buying window: a hire on a role that owns your problem, a technology added or dropped, a funding round, a regulatory deadline, a repeated visit to your pricing page. It is not a proof of intent, it is a change in probability.
- Which signals convert best?
- The ones that score high on three axes at once: recent, not visible to everyone, and close to the problem you solve. In practice, engagement signals on your own properties and organisational changes inside the team that owns your problem outperform public company news, which everyone else is already acting on.
- Why do funding rounds convert so poorly?
- Because they are public, dated, and indexed by every prospecting tool on the market. A company that raises receives dozens of near-identical messages in the same week. The signal is real, its scarcity is zero, and scarcity is what buys you a reply.
- How many signals should a team track?
- Fewer than most teams start with. Three to five, chosen because each one triggers a different message, is enough to run a signal-based motion. Beyond that, reps stop being able to tell which signal fired, and the messages collapse back into a generic sequence.
- How do you know a signal works?
- By measuring reply rate per signal, not per campaign. Tag the signal that triggered each sequence and compare after thirty to fifty attempts per signal. Anything that fails to beat your baseline sequence gets dropped, however clever it looked.