Signal-based prospecting: how to find a reason to write today

For about a decade the winning outbound strategy was arithmetic. Reply rates were low but stable, so the way to more meetings was more messages: more addresses, more mailboxes, more automation. The constraint was how much you could send, and every tool sold against that constraint.

Two things ended it. Recipients stopped reading generic mail, and the receiving providers formalised what they had been doing informally — publishing bulk-sender requirements with explicit complaint and bounce thresholds, and enforcing them. Volume became something with a price attached, paid in domain reputation.

What replaced it is not a clever trick. It is the older idea that a message works when the recipient has a reason to care about it this week, made operational by the fact that most of those reasons are now published. Companies announce their hiring, their funding, their launches and their technical decisions in public, continuously.

Signal-based prospecting is the practice of building the list from those announcements rather than from a static export. The list is smaller, it changes every week, and every account on it comes with a first line already written.

The useful test is not how sophisticated the source is. It is whether you could send the prospect the link the signal came from without embarrassment. That single question eliminates most of what is sold as intent and leaves a short, dependable set.

• Hiring posts. The most informative artefact a company publishes: written by the team doing the work, approved by someone who found the budget, and specific because it has to attract applicants. See hiring signal.
• Funding notices. Budget, mandate and a public plan to spend against, with a stated timeframe. See funding signal.
• Leadership hires in the function you sell to. New leaders replace tooling in their first two quarters more often than at any other time.
• Published technical change. A migration, a new integration, an architecture decision written up on an engineering blog.
• Expansion. A new market, a new office, a new product line — each implies work that has to be bought or built.

What is missing from that list is anonymised browsing data. It may well be accurate. It cannot be shown to the prospect, cannot be checked by the person working the account, and produces outreach whose premise nobody can defend when asked. A signal you cannot cite is a signal you cannot write from.

The most common way this goes wrong is inverting the order. A feed of signals is compelling, and it is easy to start working whatever it surfaces — which produces a list of companies that are definitely doing something, and not necessarily companies you can sell to.

Fit comes first, always. Write the ideal customer profile as a query — employee band, sector, country, stage, stack — and let the signals rank the accounts already inside it. Signals answer when. They have no opinion at all about whether.

The practical consequence is that a signal-based list is much smaller than people expect. In a given week, the number of accounts that are both a genuine fit and visibly doing something is usually tens, not thousands. That is the correct number, and treating it as a shortfall is how teams talk themselves back into volume.

A signal feed that is worked chronologically is worked from the top until the person runs out of time, which means the ordering of the source dictates your priorities. Ranking replaces that with a decision.

A relevance score is a sorting device, not a probability. It should be reconstructible — someone can look at a signal and see why it scored what it did — and it should support a floor, so the same feed can be worked as a wide sweep or a short high-signal queue without maintaining two lists.

Two signals against the same company are worth more than twice one. A raise followed by hiring against it is a much stronger case than either alone, and it is only visible if signals are attached to accounts rather than kept as separate feeds.

The signal belongs in the first line, once, in the words a person would use — and then you move immediately to the consequence you can speak to. The failure mode is the email that recites the trigger in detail and never explains why the sender is writing, which reads as surveillance rather than relevance.

Opening

One clause on the signal. Not a paragraph, and not a compliment.

Second line

The consequence you can speak to — the problem the signal implies.

Evidence

One specific thing you know, not three claims about your category.

Ask

Small, and answerable in one sentence by someone busy.

The second discipline is a deadline on the trigger itself. A funding round is a reason to write for about a quarter; a job posting for about as long as it stays open. Working an expired signal is worse than working none — it advertises that the list was built once and never refreshed, which is exactly the impression the technique exists to avoid.

Written out, the motion is short. Its value is that every step produces something the next one uses, so nothing is re-derived by hand.

• Filter the feed by signal type and score, inside the accounts your profile already qualified.
• Identify the person, not the company. Faceted prospect search by department and seniority — the address printed on a posting is usually a role address, which is the worst destination for a personalised message.
• Reveal narrowly. A reveal is a decision, not a reflex; revealing everything a filter returns is how a monthly allowance is spent on people nobody writes to.
• Enrol in a short sequence that stops the moment anyone replies, sending inside a window in a timezone you chose.
• Work the reply on the same record — through the pipeline, the quote, the signature and the invoice, without re-keying anything.

In SalesShift each of those steps is a screen rather than an integration: signals carry a type, a source URL and a score and can be pushed straight to a lead; prospect search is faceted by seniority and department; sequences stop on reply; and the deal, the contract and the invoice sit on the record the conversation started on.

The metric that matters is replies per hundred sent, segmented by signal type. It is the only number that tells you whether the evidence you are choosing is the evidence buyers respond to, and it will differ sharply between signal types — which is the whole point of measuring it that way.

Watch two counter-metrics alongside it. Complaint rate, because a signal-based list that generates complaints is not as relevant as you believe. And time-to-first-touch after a signal appears, because the entire premise decays with age.

What not to measure: total sends. It is the number the old strategy optimised, it will go down, and treating that as a problem is how teams drift back to the approach that stopped working.

If you run outbound for a specific kind of business, the recruiting, agency and developer-tools pages work the same motion through the constraints each of those has.