The pitch for signal-based selling usually arrives with an enterprise price tag attached. Intent data platforms, buying committee mapping, predictive scoring models, six figure annual contracts. It is easy to conclude that the whole category is for teams with a RevOps function and a budget line for data, and to go back to working the same static list you exported in March.
That conclusion is wrong, and expensively so. The performance gap between signal-triggered outreach and list-based outreach is the largest single gap in the outbound data. Signal-based selling teams report win rates of 33 to 41% against 18 to 25% for conventional cold outbound. Reply rates on signal-based campaigns land between 15 and 25%, against an all-campaign average of around 3.43%.
Those are not incremental improvements. They are the difference between outbound working and outbound being a cost centre that someone defends in a quarterly review. And almost none of the underlying advantage comes from expensive data. It comes from changing what triggers the send.
What a signal actually is
Strip away the vendor language and a buying signal is an observable event that changes how likely someone is to buy in the next few weeks.
That definition does two useful things. It rules out attributes, which are static facts about a company like industry, headcount or location. Attributes tell you whether someone could buy. Signals tell you whether they might buy now. A list filtered on attributes alone is a list of people who have no particular reason to respond this week, which is why it performs like one.
It also rules out most of what gets sold as intent data. Anonymised page-visit intent can be genuinely useful at volume, but it is probabilistic, it is expensive, and it is not where a small team should start. The signals with the highest hit rates are public, specific and free to observe.
The strongest categories, consistently, are leadership changes into decision-making roles, funding announcements, technology stack changes, competitor engagement, and engagement with bottom-of-funnel content on your own site.
Each works for a different reason, and understanding the reason tells you how to write the message.
The five signals worth building a habit around
A new executive in a relevant role is the single most reliable trigger available to a small team. Someone appointed as VP of Sales, Head of Operations or Finance Director has a mandate to change something, a budget conversation coming, and a period of roughly ninety days where they are actively looking at how things are currently done. They are also, unusually, willing to take meetings with vendors, because they are building a picture of the market. The window closes once they settle into the existing tooling.
Funding announcements work for the obvious reason and one less obvious one. The obvious reason is money. The less obvious one is that funding creates a hiring plan, and a hiring plan creates process problems that did not exist at the previous headcount. A company going from twelve to thirty people will break its sales process, its onboarding, and its reporting within two quarters. That is a specific problem you can write about.
Technology stack changes signal both budget and dissatisfaction. A company that just added a CRM has decided the spreadsheet no longer works, which means adjacent problems are also being examined. A company that just removed a tool has a gap and a recent, unhappy memory of why. Stack data is observable through public job ads, careers pages and site technology, and none of that requires a subscription.
Competitor engagement is the highest intent signal and the hardest to observe directly. What you can observe is proxy evidence: a competitor's tool appearing in a job description, a review posted on a comparison site, a question asked in a public community. These are lower volume but convert at a much higher rate, because the buyer has already decided the category is worth money.
Engagement with your own bottom-of-funnel content is the signal most teams have and ignore. Someone who read your pricing page twice this week is more likely to respond than anyone on any purchased list, and you already own the data. The failure here is almost always operational rather than technical. Nobody is watching, and by the time the monthly report surfaces it, the moment has passed.
The timing advantage is the whole game
The research puts a number on why this works. Teams that identify buying signals six to seven weeks earlier than competitors gain a structural advantage, because they are in the conversation before a shortlist exists.
This matters more than it first appears, given how B2B buying now works. Buyers spend only around 17% of their total purchase journey time meeting with potential suppliers, and when several suppliers are in play, time with any single rep can drop to 5 or 6% of the journey. Roughly 80% of the process happens without you in the room.
Arriving late means arriving into a process where the criteria are already written, usually by whoever got there first. Arriving six weeks early means being the vendor whose framing the buyer uses to evaluate everyone else. Same product, different position, and the difference shows up in that 33 to 41% win rate.
It also explains why signal-based reply rates are so much higher without the copy being materially better. A message that arrives during the window is relevant by construction. A message that arrives outside it has to manufacture relevance through writing, which is much harder and works much less often.
Building a signal system without buying one
The practical version of this for a team under fifty people is unglamorous and cheap. It has three parts: a watch list, a set of triggers, and a response that is already written.
The watch list is the constraint most teams get wrong. It should be small, in the range of 150 to 400 accounts, and it should be accounts you would genuinely like to win rather than accounts that match a filter. Small enough that you can monitor them, specific enough that a signal actually means something. A watch list of 5,000 accounts is a list, and you will not monitor it.
The triggers are the observation layer. LinkedIn job change alerts on your watch list contacts. A saved search for job ads mentioning your category or a competitor. Funding announcement feeds filtered to your segment and geography. Your own site analytics with alerts on pricing and comparison pages. None of this costs meaningful money and all of it can be set up in an afternoon.
The responses are what separate teams who do this from teams who intend to. For each signal type, write the message once, in advance, with the observed detail as the only variable. When a funding round fires, you are not composing an email, you are filling in one blank and sending. The reason most signal programs die is that each signal creates a small writing task, and small writing tasks accumulate until the whole thing gets skipped.
Set a rule for response time. A signal acted on within 48 hours is worth several times the same signal acted on in three weeks, because the window is the entire advantage. If your process cannot move that fast, the process is the bottleneck rather than the data.
Writing to a signal without sounding like a stalker
The message is where most first attempts go wrong, and the errors are consistent enough to be worth naming.
The structure that works is short. One line establishing the signal, one or two lines about the problem the signal usually creates, and one question the recipient can answer without a meeting. That is the whole email. Anything longer is competing for attention against a job the recipient is already behind on.
The signal line should be casual rather than triumphant. "Saw you've brought on a Head of Revenue Operations" reads as normal market awareness. "I noticed that on the 14th of August you announced the appointment of your new Head of Revenue Operations" reads as a database. Both contain the same fact. Only one of them sounds like a person.
The problem line is where the actual work sits, and it is the part teams skip. A signal tells you what happened. It does not tell the recipient why you are writing. The bridge has to be a specific consequence of the event that is plausibly on their mind. A new RevOps hire usually means reporting is about to be rebuilt. A Series A usually means the sales process that worked at twelve people is about to be tested at thirty. Name the consequence, not the product.
The question should be answerable in one line and genuinely open. "Is territory design part of what they're picking up, or is that settled?" invites a reply. "Do you have fifteen minutes on Thursday?" invites a decision the recipient has no reason to make yet. Reply rate is the metric that matters at this stage, and questions generate replies more reliably than calendar links.
One useful discipline: write the email as though you knew nothing about your own product. If the message still makes sense and still deserves a response, it will work. If removing the product leaves nothing, the signal was decoration on a pitch rather than the reason for the message.
Scoring, and why most teams overbuild it
There is a strong temptation to build a scoring model. Assign points to each signal, weight them, sum them, sort the list.
For a team under fifty people this is usually a waste of effort. Scoring models earn their keep when the volume of signals exceeds the capacity to act on them. If you are getting eight signals a week, you do not need a model to tell you which to work. You need to work all eight.
What is worth doing is a crude three tier split. Act immediately on high intent signals like competitor evaluation or repeated pricing page visits. Act within the week on strong contextual signals like a new executive or a funding round. Log and monitor everything else. Three tiers, no weightings, no maintenance overhead.
Revisit this when signal volume genuinely exceeds capacity. Building a scoring model before that point solves a problem you do not have, using time that would be better spent expanding the watch list.
The data underneath the signal still has to be right
There is one dependency that gets glossed over in most explanations of this approach, and it will quietly undo the whole system if you let it.
Signals are attached to people, and people move. B2B contact data decays at roughly 2.1% a month, which compounds to around 22.5% a year, and email addresses decay faster still at about 3.6% a month. Estimates for annual email decay range from 23% to over 40% depending on the sample and the industry. A database that was 90% accurate a year ago is likely somewhere between 63% and 70% accurate today unless it has been re-verified.
That matters more for signal-based selling than for list-based selling, which is counterintuitive. A stale list produces bounces and disappointment. A stale watch list produces something worse, which is a signal you cannot act on. The alert fires, the moment is right, and the contact left eleven months ago.
The irony is that job changes are simultaneously your best signal and your biggest data problem. The same movement that makes someone worth contacting is what breaks the record you were holding. Teams that get this right treat the watch list as a live object rather than a stored one, re-verifying contacts on a rolling basis rather than annually.
Practically, this means two habits. Re-verify contacts on the watch list at least quarterly, and treat every bounce as information rather than an error, because a bounce on a watch list account usually means someone moved and there is a replacement worth finding.
The failure modes
The first and most common is treating signals as a list-building exercise. Teams set up alerts, collect signals for six weeks, batch them into a list, and run a sequence. This removes the timing advantage entirely and converts signal-based selling back into list-based selling with extra steps. The signal is only valuable while it is fresh.
The second is over-referencing the signal. There is a difference between "congratulations on the Series A, I imagine hiring is the priority" and a message that recites four facts about the company to prove research was done. The second reads as surveillance and performs badly. Name the signal once, briefly, then talk about the problem it creates.
The third is monitoring accounts you cannot sell to. A signal from an account that will never buy is noise, and enough of it teaches the team to ignore the alerts. Prune the watch list quarterly.
The fourth is automating the send. This is the sharpest lesson from the AI SDR data, where fully autonomous agents generate volume but consistently underperform hybrid setups on reply rate and quality, and agents running without live buying signals produce reach without relevance. Automate the detection. Keep a human on the send, at least until you have proof.
What this looks like after a quarter
Expect the volume to be lower and the results to be better. A team that was sending 800 emails a month to a static list might send 200 signal-triggered emails and book more meetings, because the 200 arrive at moments when someone is receptive.
Expect the reporting to change too. Reply rate stops being the headline metric, because signal-triggered reply rates are high enough that they stop being the constraint. The constraint becomes signal volume, which means the question shifts from "how do we improve the email" to "how do we see more of what is happening in our market". That is a better question and a more durable advantage.
Expect resistance from anyone measured on activity. Signal-based outbound produces fewer touches and better outcomes, which looks like underperformance on an activity dashboard. If your team is measured on emails sent, fix that before you start, or the system will be abandoned within a month.
For teams that want the watch list, the enrichment and the sequences in one place rather than stitched across four tools and a spreadsheet, Empiraa Signal is built for this workflow, with Prospect Spark surfacing fresh prospects each month so the watch list stays live rather than going stale.
Frequently asked questions
What is signal-based selling?
Signal-based selling prioritises accounts and triggers outreach based on observable buyer events rather than static attributes like industry or headcount. Instead of working a list built in advance, the team responds to things that just happened, such as a leadership change, a funding round, or a technology stack change. Signal-based teams report win rates of 33 to 41% against 18 to 25% for conventional cold outbound.
Do I need to buy intent data to do signal-based selling?
No. The highest-performing signals for small teams are publicly observable and free: executive appointments, funding announcements, job ads revealing tech stack or hiring plans, and engagement with your own pricing and comparison pages. Paid intent data adds value at high volume, but it is not where a team under fifty people should start.
How quickly do I need to act on a buying signal?
Within 48 hours where possible. The advantage of signal-based selling is timing, and it decays fast. Teams that identify signals six to seven weeks earlier than competitors reach buyers before a shortlist and evaluation criteria exist, which is where most of the win rate improvement comes from.
Which buying signals convert best?
Competitor evaluation and repeated bottom-of-funnel page visits are the highest intent, though they are lower volume. Leadership changes into decision-making roles are the most reliable high-volume signal, because a new executive has a mandate to change something and a roughly ninety day window of active market evaluation. Funding announcements and technology stack changes sit between the two.
How many accounts should be on a signal watch list?
Between 150 and 400 for most teams under fifty people. The list needs to be small enough to monitor genuinely and specific enough that a signal from any account on it is worth acting on. Larger lists tend to become static lists with alerts attached, which removes the advantage.
Can signal-based outreach be fully automated?
Detection can and should be automated. The send should not be, at least initially. Fully autonomous outbound agents generate volume but consistently underperform hybrid human plus AI setups on both reply rate and quality, and automated sending without live signals produces reach without relevance.
