I talked to Jeanette Stepanchuk, Fractional Demand Gen & Co-Founder at Cusp GTM and former head of demand gen at Upwind Security and Anvilogic, about signal-based marketing. It's showing up near the top of marketing leaders' investment plans for 2027. Yet most of what gets written about it focuses on outbound teams reacting to the same third-party data their competitors can buy.
Jeanette's definition starts somewhere else: with what your marketing creates in the first place.
Here's what I took away from the conversation, including the parts where I still have questions.
What is signal-based marketing?
Most of what you see on LinkedIn right now is really signal-based outbound. A tool surfaces a funding round or a job change, and an SDR sequence fires. Jeanette doesn't knock that, but she points out that outbound efficiency is capped by the quality of the signals going into it.
"So my definition of signal-based marketing is when your marketing engine is designed to create first-party signals that are unique to your organization," she told me. Those first-party signals then get correlated with second-party signals from channel partners and community sources, and with the third-party signals you'd pull through a tool like Clay.
From there, you build what she calls mini algorithms around the use cases your go-to-market team cares about. When an account engages with a specific use case across several personas, the engine tells sales which play to run and who to run it on.
Put simply, signal-based marketing is a go-to-market model where marketing is built to generate proprietary engagement signals, then combines them with partner and third-party data to decide which accounts to activate, when, and with what message.
The structural work matters as much as the strategy. If your campaigns aren't named and organized by use case, the signals they produce can't be grouped by use case later. Jeanette pointed to the paid media naming convention we used together at Anvilogic as one example of that groundwork.
"A gold mine of signals"
Jeanette came up through an MQL-measured role and left it on purpose. She wanted somewhere she could build a real demand engine, and that turned out to be Anvilogic.
The MQL job left her with a story she still tells. At the end of each quarter, she went opportunity by opportunity in Salesforce, looking for any contact with a marketing touch inside the 60- or 90-day window, then told the rep the deal counted as marketing's.
Maybe the persona on the opp had opened a 15-page PDF or registered for a virtual workshop. In her words, "that definitely was not the moment that they were like, 'I'm gonna sign, like, a $2.1 million contract with you guys.'"
Her broader critique is that the model works backward from lead volume by channel. You commit to a few thousand leads from LinkedIn Lead Gen and content syndication, and buyer experience pays for it.
She watched pipeline numbers get inflated by credit hunting rather than grow quarter over quarter. She's also clear that if your board measures you on MQLs, that's the job you were hired to do.
At Anvilogic, the team ungated content, led with value, and stopped measuring MQLs. About a year and a half in, they realized the demand engine they'd built was also producing a huge volume of signals they weren't acting on efficiently.
Sales was happy and working warm accounts. What was missing were the systems and automations to move faster on them.
Do you have to throw out the MQL model?
I pushed on this, because Jeanette hears the same criticism I'd raise: isn't this just lead scoring at the account level? She agreed it's a scoring model. She thinks it's a better one, because it optimizes for surrounding a whole buying committee instead of tracking one lead's clicks.
She doesn't recommend scrapping MQLs entirely. Her example: a target account is running hot, and one contact there signed up for your newsletter six months ago and has opened it consistently in HubSpot.
That person gets a personalized approach, while the rest of the buying committee gets a broader play based on the account's signals. Keep the model as a layer, and measure marketing on something else.
That something else is a shared North Star: pipeline, built by marketing and sales together. Jeanette told me she doesn't care whether sales booked the first meeting with a cold call. For board reporting, she watches leading indicators like meetings with target accounts, how fast those meetings convert compared with before, and whether the right personas are showing up.
Dreamdata's CEO made almost the same case on this podcast a few weeks earlier. Where I still land a little differently is channel diagnostics. You need some read on Google versus LinkedIn versus events to allocate budget.
Jeanette's answer was to look at touchpoint patterns before an account becomes an SQL or opportunity, across paid, website, and content, instead of crowning one channel. I like that framing. I'm less sure it fully settles the conversation when a CFO wants a channel-level number, and I don't think either of us has a clean answer for that yet.
List everything, rank it, then organize around use cases
When I built a signal table in Clay about a year ago, I kept adding signals until almost no accounts surfaced in a 30-day window. Jeanette hears versions of that problem constantly, and her fix starts with tedious work.
Open a spreadsheet and list every signal source your marketing already produces: LinkedIn campaigns, content, newsletter, events, community, podcasts, referral traffic, and every website page down to the solution subpages. Then rank them. Without a ranking, you end up with 200 signals on a page and no idea what to do next.
The ranking comes from use cases. Which problems are product marketing, the website, and sales all aligned on? Are reps trained on them?
Once that's agreed, signals get grouped by the use case they point to, and your campaign naming is what makes the grouping possible.
I tested this with a recruiting software example. If customers tend to have 10 or more open reqs when they come in, that third-party signal marks the moment the pain is real, and it's worth latching onto. Jeanette agreed, with a caveat: how much weight third-party signals deserve depends heavily on your market.
Which signals work with skeptical, technical buyers?
Cybersecurity is where Jeanette has spent the last few years, and it's a hard place to lean on third-party signals alone. Sales cycles are long, buyers are skeptical, and trust has to come first.
A new CISO is a useful trigger, maybe for an executive introduction. It gets far stronger when campaign data shows more of the buying committee engaging with a specific use case at the same time.
Her strongest signals come from community. At Anvilogic, she built Detection Engineering Dispatch, a community built around a podcast, and nobody who signed up got prospected. But when someone attended consistently and their account was also engaging with thought leadership ads and executive content, the team knew something was happening and could reach out with something useful.
Reddit worked in a similar way. The ads there were community-led, built around the State of Detection Engineering report instead of the product. Referral traffic and engagement from Reddit threads became some of the signals she valued most, because they reflect a good experience with the brand, which compounds into branded demand.
She also brought up how Anvilogic found Omni Lab. She and a colleague had been saving and sending each other my LinkedIn posts on paid media, so when they needed an agency, we were the first call.
It's an awkward example to include on our own blog. It's also an accurate picture of how first-party signals form long before anyone fills out a form.
The minimum viable tech stack (and whether to vibe code it)
The question Jeanette gets most is whether you can vibe code all of this with Claude. You can.
Her concern is bandwidth, since most people asking are solo marketers also running content, the website, and executive LinkedIn. Every signal source needs its own API connection, and some, like LinkedIn, require an application and review process.
The practical setup she described is Clay plus an aggregation layer that pulls signals together and runs the model. She has built this in HockeyStack and mentioned Common Room, UserGems, and Dreamdata as options.
For a recent client, her team used Factors.ai, which offers a monthly plan instead of an annual commitment. HubSpot can handle part of it, but she found it limited, especially on website data.
Her vendor advice came down to budget and fit, plus a realistic note that many of these platforms are startups still building the product while you use it.
I'm more skeptical of the build-everything trend than most of what I see on LinkedIn. We did build a budget pacing app into Omni OS, and it replaced a SaaS tool that cost us about $36,000 a year. That made sense because pacing is core to how we run paid campaigns.
Rebuilding your CRM because you can is a different decision. Jeanette's version of the same worry is simple: who fixes it at 2 a.m. on a Saturday when it breaks?
Where the human stays
Correlating signals and surfacing accounts is exactly what automation should do, and every vendor above already runs on LLMs. Messaging is where Jeanette draws the line. She has Claude skills and agentic workflows that draft outreach, and she still spends hours rewriting it, because she doesn't believe a skill fully captures the point of view she's built from knowing her segment so well.
Her advice for marketing leaders is to map each person's strengths and put them in the loop where those strengths count. A content marketer with a quirky or deeply technical voice is worth more now, because their work stands out from generic output.
Final thoughts
Does signal-based marketing really replace everything you know? Partly.
MQLs still have a role in Jeanette's model, as context on accounts that are already moving. What changes is what marketing is built to produce and what it gets measured on.
Her closing point stuck with me. Buyers burned by years of MQL tactics are still recovering, and she worries about the ones getting burned by AI-written outreach now: "we're like healing a generation of B2B buyers." She thinks that's part of why big, human brand moments are landing right now.
If you want to start, the first step is the least glamorous one. Build a spreadsheet of every signal your marketing already creates, ranked against the use cases sales actually sells, and see what's been sitting there.
