← All articles

B2B Buying Signals: Which Ones Predict Replies

October 5, 2026 · 9 min read · by the 0effort team

TL;DR: The B2B buying signals most likely to earn a cold-email reply are fresh, tied to a named person, and safe to quote in the first line: a prospect's recent public post, then a recent job change. Third-party intent surges and funding news are weaker reasons to write. These are our priors, not measured findings; test on your own sends.

Most buying-signal articles rank signals by how strongly they suggest a company will buy. That's a fine question for a forecasting model and the wrong one for outbound, where you're predicting whether one human answers one email this week.

We build an AI SDR, so we read a lot of signal-driven sends. Our position: a signal is worth what it does for the first line and the reply, not what it says about the account's budget. Rank on that basis and the order changes.

Predicting a reply is not predicting a purchase

A buying signal is any observable event that makes a prospect a better target now than last month. A new VP of Sales. A job ad for your category's buyer. A conference talk on the exact problem you solve.

An account can be deep in a buying process and still ignore you, because the person you emailed isn't on the committee. A person can also reply warmly with no purchase in sight. So we'd argue three things govern reply likelihood: freshness (how recently it happened), attachment (a named person or only an account), and referenceability (whether you can mention it in line one without sounding like you've been watching them). Standard signal rankings score none of the three.

Seven categories of buying signal and where each comes from

Seven kinds cover most of what outbound teams use. Shelf lives are our assumptions, not measured benchmarks; the testing section below replaces them with your own numbers.

The two rows with the strongest "intent" reputation are the two we'd never quote back to the prospect.

Which signals tend to predict replies best, and what do they have in common?

Our prior: the person's own recent public actions, then role changes. We hold the order loosely; the reasoning matters more.

Both are recent, so the email lands while the topic is still on the prospect's mind. They attach to the person you're writing to, so the email reads as addressed to them rather than to a job title. And they were published for an audience, so mentioning them is expected rather than eerie.

Hiring posts rank third, and only on referenceability: they're a public ad, so citing one is safe. They attach to the account and stay true until the role closes, so they say little about whether this person is paying attention now.

Someone three weeks into a new job is reviewing tools and inheriting vendors. Whether that converts for your offer is an empirical question. A public post gives the first line something true to hold: "You wrote last week that your SDRs lose mornings to list hygiene" can only be a message about that person.

Which popular signals look strong on paper but are weak reasons to write the email?

Two, and our reasoning rests on how the data works, not on a lift statistic.

Third-party intent surges. Topic-surge data is typically modelled at the company or IP level. It says someone at an account, or on its network, read more than usual about a subject. It doesn't say who, whether they hold buying authority, or whether they were a student or a competitor. Treating it as proof that your named contact is in market is the core mistake. It can rank which accounts to work first, but it's a poor reason to write "I noticed your team is researching X."

Funding announcements. They change budgets. They're also public the instant the release lands, so every vendor emails the same founder the same week. Strong signal, crowded inbox; only a test shows which wins.

One caution for everything here: reply rate is not positive reply rate, and neither is meetings booked. A signal can lift total replies while mostly producing "not interested" and unsubscribes. Watch positive replies per hundred sends, with unsubscribes beside it. When we say a signal "works," we mean positive replies.

Person-level signals go stale in days to weeks

Person-level signals expire faster than most workflows can act on them. Our rule: an event tied to a person's attention (a post, a talk, a role change) decays in days to a few weeks, because the first line stops being topical. An event tied to a business state (an open posting, a tool in the stack) decays slowly because it stays true. The first group needs a send within days of detection. The second can wait for a good moment.

That asymmetry is why batch research hurts: with monthly list-building, every person-level signal arrives pre-expired. A stale signal is worse than none. "Congrats on the new role" sent three months in tells the prospect your information is old and your process is automated.

Stacking signals helps less than it sounds

Requiring several signals at once helps less often than it sounds. Many signals aren't independent: a funding round tends to precede hiring posts, which tend to precede new-leader announcements. Counting those as three signals counts one event three times. And stacking shrinks your list, which makes the rule slow to evaluate.

We'd start with one fresh, person-level signal plus a firm ICP fit, and test a two-signal rule only once the one-signal version has enough sends to measure.

How do you reference a signal in a cold email without sounding like you're watching the prospect?

Ask whether the prospect would expect you to know it. Things they or their company published pass. Things inferred from their behaviour fail.

Pass: "You posted about moving outbound in-house last month." Pass: "Your Head of RevOps ad mentions deliverability." Fail: "I noticed your team visited our pricing page." Fail: "Our data shows your company is researching AI SDRs." The failing lines reveal that you've been observing, the one thing a cold email can't afford to say.

The fix for a failing signal isn't better phrasing. Keep it in your prioritisation and out of the email: use it to decide who and when, then write line one from something public, in one clause, tied to the ask.

How do you test which signals predict replies in your own market?

Don't trust anyone's benchmark. Vendor-published ones typically come from the vendor's own customers and copy, so treat them as marketing until the sample is described. Run your own comparison:

1. One signal, one comparison. Contacts with the signal versus same-ICP contacts without it, same sequence, same copy.
2. Assign by contact, mixed across mailboxes and send days. Otherwise a deliverability gap between mailboxes masquerades as signal lift.
3. Count positive replies per 100 sends, with unsubscribes and total replies beside it. Skip open rate: tracking pixels get pre-fetched by mail clients and scanners, a problem our deliverability checklist covers.
4. Fix the stopping point in advance. Peeking on day three and calling a winner ships noise.

How many sends? Assuming a two-sided 5% significance level, 80% power and a normal-approximation two-proportion test, these are parametric calculations, not findings:

Positive-reply rates are lower, so it gets worse: 1% → 2% needs about 2,300 per arm. Most teams can't resolve a one-point gap on a signal that fires 150 times a month, so test big hypotheses and treat small differences as unproven.

What should you demand from an autonomous AI SDR that acts on signals?

We'll keep product claims to what 0effort says publicly: it sources buyers, writes and sends every email and LinkedIn touch, answers replies over email and phone, and books meetings. We won't publish a signal-feed list or a lift figure here; the argument above says you shouldn't accept one from us anyway.

What we'd ask of any autonomous system, ours included:

• Provenance on the contact record: which signal, when detected, source URL, so a human can check line one in ten seconds.
• Freshness gating: stale signals dropped or downgraded rather than sent anyway.
• Signal-level outcomes: positive replies, unsubscribes and meetings by signal type, so the test above is a query, not a project.
• No quoting of inferred signals: modelled intent can rank accounts; it shouldn't write sentences.

Automation helps most on freshness. Nobody can watch thousands of profiles and postings daily and still write within the week; a machine can, which is the intent stage of the loop in our cold email guide. It does nothing for measurement unless it records what it did.

FAQ

What is the difference between a buying signal and intent data?

A buying signal is any observable event suggesting a prospect is a better target now: a job change, a post, a hiring ad. Intent data is one kind, usually third-party and modelled from browsing across a network. It's generally account-level; many other signals attach to a named person.

Are third-party intent data providers worth paying for if you only do cold email?

Probably not as a first purchase. The data is typically account- or IP-level, so it can't say which person to email, and you can't mention it without sounding surveillant. Test it against a free baseline (ICP fit plus a fresh public signal) first.

How many buying signals should you require before contacting an account?

Start with one fresh signal, ideally tied to a named person, plus a firm ICP fit. Requiring two or three cuts volume sharply and often counts one event twice. Test a two-signal rule once your one-signal version has enough sends to measure.

Is a funding announcement still a good trigger for outbound?

It's a real trigger and a crowded one. The news is public immediately, so competitors email the same people the same week. Use it for prioritisation, pair it with a person-level signal, and judge it on positive replies per 100 sends in your own test.

Can you use buying signals for outreach to EU prospects under GDPR?

Yes, with care, but a signal is not an exemption. Anything tied to a named person, even if publicly visible, is personal data under Art. 4(1) of Regulation (EU) 2016/679, which needs a lawful basis under Art. 6(1) and, when you collect it from a source other than the person, an information notice under Art. 14. Whether you may send the unsolicited email is a separate question under Art. 13 of the ePrivacy Directive 2002/58/EC, which member states implement differently, so check with counsel (all four sources accessed 2026-10-05). Our GDPR guide for EU cold outreach covers the lawful basis in more depth.

How many sends do you need before you can trust a difference in reply rate between two signals?

It depends on the baseline and the gap. Assuming a two-sided 5% significance level, 80% power and a two-proportion test, 5% → 10% needs roughly 430 sends per arm and 4% → 5% around 6,700. Smaller gaps are hypotheses, not results.