What signal-based selling actually is

Search the phrase and every page gives you roughly the same definition. Fit tells you who could buy. Signals tell you who is buying this week. Stop working a list built once a quarter, start working the accounts where something just happened.

The first half of that is fine. The second half is a claim about intent, and no publicly observable signal measures intent. A funding round, a job posting, a new executive, none of them is a company deciding to buy something. Each one is a change in the company's situation, and those are different things.

That distinction sounds pedantic until you watch what it costs. A team that believes signals identify buyers treats every signal as a reason to act, gets a low hit rate, and concludes signals do not work. A team that knows what a signal actually is gets something narrower and much more useful out of the same data.

What a signal is evidence of

Three of the most common signal types have real research behind them. Read what each one actually measured.

Job postings. Gutiérrez, Lourie, Nekrasov and Shevlin, writing in Management Science in 2020, found that changes in a company's online job postings are positively associated with its future growth in sales and earnings, that investors treat posting changes as new information, and that the effect is stronger for growth hiring than for replacement hiring (Management Science, 2020). That is a measurement about the company's trajectory. It is not a measurement about whether the company will buy anything from you.

Leadership change. Intintoli, Serfling and Shaikh, in the Journal of Financial and Quantitative Analysis in 2017, matched customers with their suppliers and tracked what happened after the customer replaced its chief executive. Suppliers lost substantial sales to that customer, with larger losses where the departing executive had been entrenched (Intintoli, Serfling and Shaikh, 2017). The measured effect runs in the opposite direction from the way the trigger is usually sold. It is about incumbent vendors losing ground, not about new ones winning.

Funding. The cleanest of the three as a fact, and the weakest as a conclusion. A round is public, dated and unambiguous, and it tells you money arrived. It does not tell you the money is being spent on your category, and capital has concentrated enough in recent years that the same headline number buys a much smaller team than it used to. What a round actually tells you works through that with the figures.

Put those together and the honest version of the category definition is less exciting and more usable. A signal is a public, dated event that makes a company worth looking at now rather than later. It changes the order of your list. It does not tell you the answer.

The research the category leans on says something else

The strongest argument for timing is a buyer-side one, and it comes from 6sense's 2025 Buyer Experience Report, a survey of nearly 4,000 B2B buyers across North America, EMEA and APAC with a median purchase between $200,000 and $300,000.

Three findings from it matter here. Ninety four percent of buying groups ranked their preferred vendors before contacting any of them. Ninety five percent of the time the winning vendor was already on the Day One shortlist. And four out of five deals were won by what the report calls the pre-contact favourite (6sense, 2025).

Now look at how that gets used. Paraphrasing one of the pages that rank for this category, the finding is rendered as proof that the first vendor contacted wins about 80% of deals, so the job is to respond to signals within minutes.

Those are not the same claim. The report says the winner was already the favourite before anyone made contact. It says nothing about who reached out first. Being the favourite in advance and being fastest on the draw are different positions, and the research supports the first one.

That flip matters because it changes what speed is for. If the win goes to whoever answers a trigger fastest, the game is reaction time. If the win goes to whoever was already credible when the shortlist got written, the game is being relevant during the months of research that happen before a seller hears about it. The same report puts the point of first contact at 61% of the buying journey, moved forward from 69% the year before, which is roughly six to seven weeks earlier. Buyers do reach out sooner than they used to. They still do most of the work first.

Gartner has measured the space this leaves a seller. Across an entire purchase, buying groups spend about 17% of their time meeting all potential suppliers combined (Gartner). Forrester's 2024 survey of buyers found 92% start the process with a vendor already in mind and 41% with a single preferred one (reported by Digital Commerce 360).

So timing is genuinely the constraint, which is the argument for working this way at all. But the thing timing gets you is a place on a list that is being written without you, not a race won by the quickest reply.

An ordering problem, not a list problem

The most common mistake with this approach is using signals to build the list instead of to sort it.

It is an easy mistake, because a lot of the tooling is sold that way. Feed in a signal type, get back companies where it fired, work the result. The trouble is that a signal on a company you could never serve is not an opportunity, it is a distraction that arrives with a reason attached, which makes it more convincing than ordinary noise rather than less.

The order that works is the other way round. Decide who you can serve, then watch that set and let signals tell you which of them to look at this week. That puts a sharp ICP upstream of everything else, and it means the signal is answering when, never who.

There is a second reason to keep the set fixed and the order moving. Research time is the scarce resource in small-team prospecting, and the arithmetic on that is unforgiving. A monitored set you chose deliberately turns research into reading what changed. A set that reshuffles every time a new signal type fires turns it back into starting from zero.

Three things that have to be true first

A signal is only usable if you can trust three properties of it, and each one fails quietly in a way that looks like a working system.

It is attached to the right company. Funding coverage in particular attributes rounds to whoever appears in the headline, which is regularly the investor, the publisher, a former employer, or a company that happens to share a word with the product being described. Who raised the money is not who is in the headline covers the shapes this takes and why a confidence score on the match does not catch any of them, because entity resolution and role assignment are separate questions.

It carries the right date. Every signal has two dates, the day the event happened and the day monitoring saw it, and they are frequently weeks or months apart. Job postings are the worst offenders here. The date on a signal is not the date of the event explains why that gap exists and why a freshly detected signal can be an old one.

There is enough of it to read. Signal availability is not uniform across company sizes. Below a certain headcount most companies produce nothing publicly readable for long stretches, and the silence is mostly real rather than a gap in detection. What buying signals look like by company size has the shape of that. If your ICP sits in the quiet range, signals will sharpen your ordering far less often, and knowing that in advance is better than concluding the tooling is broken.

Any of those three failing produces the same experience. Events arrive, they look legitimate, and acting on them wastes time in a way that is hard to diagnose because nothing errored.

Where detection stops and judgement starts

The direction the category is moving is toward closing the loop, where a detected signal fires an action without anyone reading it. Detection and action in one system, no human in the middle.

The three failure modes above are the argument against that, and they are not exotic. Wrong subject, stale event, thin coverage. A person reading the signal catches all three in seconds, because the checks are obvious once a human looks. An automated path catches none of them, and the mistakes it makes are addressed to real people at real companies under your name.

There is a narrower version that does work. Detection, ranking and assembling the evidence are mechanical, and they are where most of the hours go. Deciding whether this particular event is worth anyone's attention is a judgement, and it is fast when the evidence is already gathered. Machines are good at the first part. The second part is the part worth keeping.

What it does not replace

Signal-based selling does not qualify anything. An account that just raised, hired and changed leaders can still have no budget for your category, no authority in the seat you reached, and no problem you solve. Timing tells you the door is open, not that there is a reason to walk through it.

It does not replace an ICP. It depends on one, and it degrades exactly as fast as the ICP is loose.

And it does not produce a guaranteed outcome. It moves you from working a list in the order it was exported to working it in the order things changed, which is a real improvement and a modest one. Anyone selling it as more than that is selling the version with the intent claim still in it.

The practical shape of it

Pick the set of companies you could genuinely serve. Watch that set for public, dated events drawn from publicly available and licensed business data. Weight what arrives by type and by how recent it truly is, not by when you first saw it. Read the top of the resulting list, decide which ones deserve a conversation, and reach out from your own tools with the specific event in hand.

That is the whole method. The signal types worth knowing and how to read a job posting like a buyer go deeper on the reading part.

Leadalise exists to do the mechanical half of that. It watches a set of accounts you choose, checks them for these events, daily for priority accounts, resolves which company each event is really about, and puts the ones that moved at the top of the list with the evidence attached. How many accounts a plan covers is on the pricing page.