Signal-based outbound: which buying signals actually work?

Who this is for: sales and marketing leaders at B2B companies deciding what to build outbound around, and anyone setting up signals in Clay or a similar tool.

* Every example and result in this post comes from my work at previous agencies, before Thresh & Winnow. Clients are anonymised.

Why do so many signal-based campaigns feel the same?

Over the last 18 months, signal-based outbound has turned into a set of templates. Most teams track the same few things: job changes, funding rounds, hiring. If you lead a sales or marketing team, you’ve seen the result in your own inbox: you can usually tell from the first line which signal the sender was tracking.

Those signals can still work. Whether they do depends on context: who you’re selling to, what you’re offering them, and whether you have proof that fits the moment. You can only choose good signals once you understand your value proposition, the personas in your ICP (Ideal Customer Profile: the companies and people most likely to buy), what those people are dealing with, and why they might buy. The signals come after that work, never before it.

Where do you start when choosing signals?

With what has already worked. Before we track anything, we ask:

• Have you tried outbound before? What happened?
• Which deals have you closed, including the ones that came in through inbound?
• What do those customers have in common: size, stage, market, the problem they bought for?

The themes in those answers point to the best-fit ICP, and the ICP decides the signals.

Your own team is the best source. Your sellers and founders know which customers stuck and why, often from years in the industry. A signal that makes sense on paper but not to the people who sell your product every day is usually wrong.

Why aren’t standard list filters enough?

Conventional prospecting tools such as LinkedIn Sales Navigator and Apollo filter by industry, size, location and job title. They’re a good starting point, but in most cases you’ll have to refine the list further to find your ICP.

At a previous agency, before Thresh & Winnow, I worked with a safety technology company that sold to underground mines. The buyers were geotechnical engineers, and the product only worked underground: not in open-pit mines, and not in coal mines, where its electronics weren’t allowed near flammable gas. The client knew this well. The lists being used at the agency didn’t reflect it yet, because filtering properly takes time.

The best you can do in a conventional tool is a title search plus a keyword search, for example in Sales Navigator:

("underground mine" OR "underground mining" OR "UG mine" OR "UG mining") NOT (coal OR surface OR "open-pit" OR "open pit")

That finds geotechnical engineers who mention underground mining and drops anyone who mentions coal, surface or open-pit mining. It’s an approximation. A profile can match because of a job someone left years ago.

So I worked backwards from the mines instead:

1. Built a list of underground mines in a spreadsheet, with how long each one had left to operate. A mine near the end of its life won’t put five figures into new equipment for an operation that’s winding down, so I only kept mines with ten or more years left.
2. Searched for engineers at those sites, using the same kind of keyword string with the names of good-fit mines.
3. Checked annual reports for injury and fatality reports, where the case for safety investment was strongest, and for recent projects at each mine to reference in the outreach.

At the time this was manual. Over nine months, the campaign booked 90+ meetings with mine engineers.

Today the same layering runs at scale. Tools like Clay and Freckle can check that the keywords sit in someone’s current role rather than a past one, pull mine and project data into the same table, and qualify every row. Sales Navigator and Apollo give you the starting list. The qualifying is where the list becomes yours.

When does a funding round work as a signal?

When you have something worth saying to a company that has just raised.

Working with a growth advisory for early-stage health tech companies, the first job was defining “health tech”. Crunchbase’s industry filters couldn’t separate a product company from a services firm, or a young startup from one that no longer needed advisors. The client defined it for me: companies selling health tech products, still operating, and either founded in the last five years without funding or within three years of a round no later than Series A. I pulled a wide list and qualified every company against that definition with Claygent, Clay’s AI research tool.

Then I layered on the signal: a raise in the last 12 months, a sign the company was entering a period of growth. The email congratulated them on the raise and brought in the firm’s experience. Breaking into large, established hospitals is hard for a young health tech company, and the firm’s senior advisors could open those doors and shorten the time to close. They had case studies with numbers, from companies at the same stage.

On the best campaign, a few hundred emails, 1 in 4 messages got a positive reply.

I then wrote to Venture Capital (VC) firms with health tech portfolios. Each email named at least two portfolio companies that had raised recently, to show I’d done the research, and explained how the firm could help their portfolio companies. 1 in 6 of those got a positive reply.

The raise worked because the case studies matched the moment. Track the same signal without relevant case studies and you’re congratulating someone on their raise with nothing of substance to offer.

What does signal layering look like?

Layering means combining signals so each one narrows the list or sharpens the reason to write. In the health tech campaign it ran in three steps: fit, then timing, then proof.

A campaign for a sports sponsorship analytics company shows the same idea with signals specific to one market. When a team signs a sponsorship deal, there’s usually an announcement quoting someone senior in marketing or sponsorship about what they want from the partnership. I pulled that quote and the objective it named, and wrote to that person about measuring the value the team was delivering to its sponsors, which helps a team negotiate better deals. For brands, I counted how many major leagues they sponsored in (NFL, NBA, MLB, NHL, MLS). Brands with deals across several leagues were spending heavily on sponsorship and had the most to gain from optimising it, so they became first-tier accounts. It performed well because the signals came from how that market actually works.

Can a campaign have great signals and still fail?

Yes. For a company selling authentication to software teams, I built one of the most complex signal setups I’ve worked on. I looked for user growth (annual reports, website traffic), new products, new country expansion, and mergers and acquisitions (M&A), from quarterly reports, news and company pages. Every layer pointed to companies whose user numbers were about to grow.

It didn’t perform. The client had done well with mid-market companies and wanted to move into enterprise, but every case study they had was mid-market. I raised it more than once. The push into enterprise went ahead anyway, and the proof never matched the buyers I was writing to.

The lesson: signals decide when you write to someone. The offer and the proof decide whether they reply. No amount of layering fixes a mismatch between the two, and a clear view of where you’ve actually won matters more than a clever setup.

How do you check a signal is right before it goes out?

A signal is only useful if it’s true and relevant. For the raise campaigns:

• Relevance rule: the raise had to fall within 12 months of the date the email was sent.
• Sources required: the research prompt had to return a link to its source for every raise.
• Spot checks: I checked a sample by hand to make sure the AI step wasn’t making anything up.
• A fixed format with fallbacks: when a signal goes into the email, the line has to read correctly every time. Mine was simple: “Congrats on your seed raise in July.” The month and stage came from the research. Seed to Series A rounds were named; anything outside that bracket, or in an unusual format, fell back to “your raise”.

That’s how a personalised line can run at scale without embarrassing anyone.

What should a rep get with each account?

The research behind the signal shouldn’t stop at the email. Whatever your reps need to know goes to them, so they arrive on the call having done the homework.

For the health tech campaign, that meant the last funding round and its date, company size and who I’d spoken to. For the VC firms, it meant information on every relevant health tech company in the firm’s portfolio: when each one last raised, at what stage, and any new Chief Medical Officer, VP of Sales or other C-suite hires.

None of this is manual. Replies the sequencer marks as positive are sent into Slack through a webhook or API. The signals you track are pushed into the CRM, and an orchestration tool such as n8n pulls the relevant research from Clay or your CRM into a short doc, shared by Slack or email with whoever takes the meeting.

How do you know which signals are working?

Treat every signal as a hypothesis. Split campaigns by signal and by hypothesis, so each one tests a single idea about the buyer, then compare positive replies and meetings across them. The more you learn about the market, the more the signals change. Put more behind what performs, rework what doesn’t, and be ready to drop a signal you liked on paper.

Key takeaways

• Understand your value proposition, your personas and their context before choosing any signal.
• Start from closed deals and your team’s knowledge, not a template.
• Use conventional tools for the starting list, then qualify it further: fit, then timing, then proof.
• Signals show you’ve done your research, but your offer and case studies have to tell the same story. Together they show you understand the person you’re writing to.
• Check every signal is accurate and relevant. If it goes into the email, write it to a fixed format with fallbacks.
• Split campaigns by signal and hypothesis, pass the research to your reps, and build on what performs best.

If you’d like signals built around your own buyers, here’s how our signal-based outbound works.