Signal Is Having Its Moment. Don't Let It Become the New MQL.

A few weeks ago I said the word of 2026 even if we're only half way through the year, was CONTEXT.

But there's a second word coming up fast: Signal.

Not new. Behavioral signals, intent signals, buyer signals, this vocabulary has been around for years. What's new is how many people are suddenly excited about it, and how fast "capture a signal" gets wired directly to "send an outreach."

That second part is where I want to slow things down. I want to build a case that when you want to use signals right you need ... Context.

The promise is real

Let's be fair to the concept first. Signals — job changes, content consumption, hiring patterns, tech installs, review site research, pricing page visits — genuinely tell you something about what's happening inside an account. AI has made it possible to collect a lot more of them, at a lot more detail, than any human team could ever process by hand. That part of the hype is earned.

And there's a real discipline here, even if it often gets reduced to a caricature. The lazy version goes: bundle a few third-party data feeds, add some AI scoring on top, call it a strategy.

Here's the thing: one signal on its own tells you almost nothing reliable. And not all signals are the same. You can basically split them in two. On one hand, signals that relate more to potential change at the account level — even if the signal is captured from the behavior of one person, and that signal is personal and emotional, it can still lead to change at the account level. Think of increasing consumption of very specific content from different stakeholders, all working at the same account. The other type is simpler, cues you can use to personalize a marketing outreach, both the content and the timing — a promotion of a key stakeholder in the buying group, for example. The problem is that most people lump all of these signals into one bag.

Let's talk about the account-level signals for a moment.

The value only shows up once you run enough of them, across enough accounts and/or people, over enough time, through a model trained on real outcomes — did this pattern of behavior actually turn into pipeline or not. Vendors like 6sense and Demandbase built an entire category around exactly this: not counting engagement points, but training on what a buying process that converts actually looks like, from start to finish, across a whole buying group, and comparing every live account against that pattern.

That's a real discipline, not lead scoring with extra columns. I don't want to undersell it, especially for the complex, multi-stakeholder deals most of our clients work.

Making use of these kinds of solutions comes with a minimum threshold, though. It isn't for everyone, and that's worth saying out loud.

Buying one of these platforms and actually making it part of your GTM takes a certain operational size, maturity, and budget to begin with. It takes patience too, which rules out most startups and scaleups still figuring out their motion. And it takes a market, a geography, a category, where the underlying signal data is actually reliable.

Caution: The knee-jerk move — using naked (intent) signals without relying on the predictability of an AI engine, and telling sales reps to use that one signal as a reason to reach out — is not a viable first step.

Which brings me to the point I actually want to make: it's not because you don't have the infrastructure to scale it, that starting out by using a single signal as a trigger is a step in the right direction, let alone the right reflex. It seldom is.

Where it goes wrong

So let's talk about those looking to do something with signals who aren't using the revenue platforms mentioned above — including the individual "personalization" signals I just talked about.

Here's what I keep seeing on LinkedIn: more and more "thoughtleaders" turning signals into scripts. They're talking to companies not yet ready to go for a predictive revenue platform like the ones I mentioned above. For that audience, signal-based marketing becomes just another alternative to what they tried to achieve with "inbound": a way to collect names and create MQLs, only with newer language attached to it.

What this school of thought is still convinced of is that these activities are a genuinely effective way to produce revenue. In several blog posts I've written — and referenced research from people like Kerry Cunningham on— the fact is that they aren't. The entire inbound practice, for all its promise, hasn't yielded the results. It optimized vanity metrics.

The formula is always some version of: signal fires, here's the exact message to send, book the meeting. As if the only acceptable outcome of a captured signal is a contact moment or a reply to an outreach. So the challenge gets translated into: we need more signals, because the signal data we have today isn't doing the job. I'd say companies have way more first-party signal data than they realize, but due to technical constraints, they can't use it. That said, I do agree that the elephant in the room is that most companies lack third-party signals altogether.

The solution? We do indeed buy signal data. And yet, we still find out it isn't yielding the results we wanted.

The thing is, this isn't really a signals problem. It's a context problem, but people are calling it a signals problem.

Side note: A signal picked up without context — without knowing where that contact sits in the buying group, or where the buying group sits in its journey — doesn't just underperform. It can even work against you.

Take a job change of a decision maker. Early in a buying journey, that's a genuinely useful trigger. New person, new priorities, worth a conversation. The same signal, treated the same way, late in a journey that's already tied to an open opportunity, can do real damage. Cold-reaching a stakeholder to "re-engage" them, because a tool fired a signal on its own, while they're already three calls deep with your AE, doesn't move the deal forward. It confuses it, or worse, it contradicts what your own sales team is doing at the same time.

Multiply that across many signal types and hundreds of contacts and accounts, and you see the issue isn't the signal. It's using it without any governance around it.

The nuance I'd add

None of this means every signal needs a model and a ton of data before you're allowed to touch it. As I pointed out, some signals just offer a reason to personalize a touch point — especially in named-account environments, where account managers need to make contact regardless just to build a relationship.

Not using any of the signals, and relying only on an AI engine to engage, would just swap one overcorrection for another.

Some individual signals are strong enough on their own to justify acting right away. But contrary to my named-account example, that doesn't always mean establishing a direct human contact — it could just as well be the start of a targeted awareness ad stream.

For those already running the revenue platforms mentioned above, like 6sense and Demandbase, this also doesn't mean neglecting signals and doing nothing with them in the meantime either. Again, in some scenario's one particular signal but certainly not all, can be valuable. Most of the signals just should serve to feed a decisioning engine.

Also, that decisioning shouldn't be rules-based. Rules-based means you're only working with a narrow set of structured data — a field value, a score threshold, a yes/no flag. AI-driven decisioning can pull in everything, structured and unstructured, to add just enough context to one signal (who is this, where do they sit in the group, what else is already going on) before deciding whether to act and how. Same predictive foundation underneath. Just applied at the speed a transactional deal needs.

There's no conflict here. One predictive foundation, applied two ways: aggregating data and signals over time for the full account — using the tooling just mentioned, if you want to scale the effort on the one hand. Or using very specific singullar signals to personalize a reach out (especially in existing client relationships to help uncover upsell, cross sell etc. Using singular signals also could make sense in transactional environments or when the buying group is small enough that one trigger can mean something on its own.

Doing all of this together sounds great on paper, but for many, if not most, this is "nec plus ultra" territory. Automated decisioning is in a very limited way doable already for those with the right infrastructure, and it will only keep getting better over time, so it can be applied more broadly. But we're still far from a situation where every marketer has the ability to scale and automate the entire decision-making process — the message, the channel, the timing — and have all of that generated and selected 100% by AI end to end. And honestly, I'd argue we'll never fully get there. If every competitor chasing the same buyer runs the same automated decisioning against the same signals, you end up with everyone converging on the same "optimal" message at the same moment. That's not differentiation, that's a race to sameness.

Two reasons this holds. First, the data isn't there yet, and I doubt it ever will be completely — from the perspective of a vendor who will always have to work with incomplete information, no matter how good they are at acquiring third-party data. What we have is still far from an exhaustive view of a buyer, however many signals you stack on top of each other. Second, and this one won't change with more data: the human, emotional side of engagement is what actually creates differentiation. It's a side that AI tends to underserve, and one that deserves more attention. The behavioral psychologists amongst us should see this as an opportunity.

The piece nobody's building: air traffic control

Here's the part almost every signal-to-action pitch skips: even a signal with proper context can still clash with something else happening in the account. Too many messages sent too fast, acting on every signal that fires, is the first and easiest mistake to spot. But it goes further than just volume. A reasonable "re-engagement" trigger can land in the same week as an open, sensitive support ticket. A "buying group is expanding" signal can fire while a rep is mid-negotiation on price. Each action on its own might be fine. Together, they feel like chaos to the customer.

That's not a decisioning problem, and it's not really a context problem either. It's a coordination problem. You need something sitting above all the individual signal-to-action logic, asking: does this conflict with anything else already running on this account? Some people call this air traffic control, and I like the image.

I should be honest that the example above is a simple one, and to some extent even a rules-based system can catch it. Don't send this if an open support ticket exists, don't send that if a deal is in a certain stage. That's a fixed set of if-then logic, defined upfront on structured fields, and it works fine for the cases you thought of in advance. I only used the example to make the idea concrete.

The harder problem sits somewhere else. Now that we're starting to communicate with several contacts in the same buying group at the same time, in an automated and coordinated way, using aggregated signals as triggers, the response or action of one member can change what the next best action should be toward the others. Someone replies, someone forwards an email internally, someone goes quiet after being vocal, and that shifts the picture for everyone else in that group. You can't lock that in a rule you wrote six months ago, because you didn't know which member would move first, or how. That's the level of air traffic control that needs something watching in real time, not a rulebook written in advance. Someone, or something, needs to see the whole picture, not just clear each flight for takeoff on its own.

Very few platforms, if not none, are built for that today. Most are built to make the signal-to-action part faster, which is exactly the part that most needs slowing down. Either it's done by importing signals into a marketing automation system with predefined nurture or engagement tracks, or it's just a collection of separate tactics scattered across different platforms — an SDR AI here, an email tool there. In both cases it will hit a limit, hurt sales cycles, and even hurt effectiveness and reputation.

Where the predictive layer actually pays off

None of that is an argument against the predictive and real-time decisioning layer. The point I'm making is that we shouldn't expect it to run the whole show alone. Even short of full automation, with a human still checking or overriding the output, it earns its place fast. And we'll see a lot more end-to-end decisioning over time, that much is for sure.

Aggregated, predictive signals already power real internal processes on both sides — next best action for a sales rep deciding who to call today, or a marketing platform deciding what to trigger next. That's not theory. It shows up directly in pipeline.

But we still have a long way to go, and human oversight — governance and control — remains critically important. For many, the air traffic control function today is a challenge that still requires a lot of human intervention (perhaps supported by better reporting).

On a related note: we're currently working with a brand new martech vendor on the launch of a marketing automation platform that's AI-native at the core and actually understands buying groups, not leads. But also is AI and not rules based, and has air traffic control built in. The founder, Jon Miller is the former co-founder of Marketo. It's genuinely exciting, and if you want an early look at what this could mean for your own stack, happy to arrange a demo.

And what is the story when you only have your "old" marketing automation platform to work with?

If you're not ready for a platform upgrade and just want to add signals to what you already run, there's still plenty you can do. It takes structure, though, and the discipline not to build a forest of signal-to-response trees that nobody can maintain six months from now. You can use signals to more productively power engagement methods you already have in place — an ad platform, an SDR AI, whatever fits your stack. But it needs governance and control wrapped around it. As long as the volume stays manageable, that's exactly the kind of thing a managed service, like the one we run, can hold together for you.

That said, we'd caution you against going overboard following the thoughtleaders pushing you to build dozens of signal-to-action workstreams.

And to be clear, going back to the point above: we genuinely think a fully automated engagement system is a pipe dream as well, for the same reason. The human touch remains the one thing that actually differentiates one approach from the next. Not all of the steps will be digital either.

Final conclusion: buzzword bingo — there's predictive, and there's predictive

Up until now, we've used the term "predictive" to describe knowing what the next best action is within an engagement stream. But that's a pretty vendor-centric way of defining marketing. It's important — once you have an account in your crosshairs, you do indeed need to understand how to engage. And the engine telling you to do that well has to use a plethora of (historical data). But at the same time something else needs to go on as well, because there's a whole other area before you have them in your crosshairs: when they're in market but don't want to engage with vendors yet. That's where marketing also has to play a role. And predictability matters here too — just a different kind. This one is about understanding the next move of one account in your ICP, and of the members of the buying committee.

This is about marketing actually showing up at the right moment, with the right message, in front of the right stakeholder. That's a separate discipline, and it's older than any of this AI tooling. SiriusDecisions once had a name for it: engaging your target based on their presence near the watering hole, being where your audience naturally gathers, at the moment they're actually there, not just knowing that a herd is moving somewhere.

Predictive signal aggregation tells you the herd is moving. It doesn't, by itself, put marketing at the right spot on the trail with the right message when one "animal" is arriving.

That's the gap worth opening up next: how marketing actually builds that watering-hole presence in a signal-driven, AI-native world, and where that job sits next to the decisioning and air-traffic-control pieces described above.

This is not something any technology today will solve on its own. You still need to understand the way your buyers actually buy from you, the dynamics inside the buying group, the needs of its members across the journey. The more you develop that understanding, the better you'll be able to engage. But it starts there, not with a tool. From that understanding, you'll know better which signals actually matter, which ones can lead to what kind of action, and for a while you can even run and govern all of that manually. Over time, though, you'll want to scale the effort and bring in more machine support, through a new breed of martech applications like the one I mentioned above.

There's so much more to explore. But the human in marketing is not dead. On the contrary.

Recommended steps

1. Invest in Buyer Insights: do the research and document how your buyers buy — the target audience (TAM) and buyers of the solution you've built for their business challenge.
2. Define the ICP segment of this TAM, but above all, maintain it.
3. Augment the static ICP with enriched data (demographic, psychographic, you name it), but also including signals.
4. Use the Buyer Insights to make sure you understand the triggers that start a buyer's journey, and the triggers that, during that journey, make certain members of the committee take certain actions.
5. Structure these signals and assign them a level of importance.
6. Build engagement treatment streams using only the critical signals to aim for a contact moment, and use all the others to help automate less intrusive, multi-channel engagement.
7. Make it easy — very easy (!!) — for your buyers to contact you once THEY are ready.
8. Monitor the engagement, tweak it, change it.
9. Use most of your existing martech stack, and learn to do it right.
10. Investigate, and more importantly, be patient. Don't rush to buy more tech — most of what you'd need to do this right is still being built.