GTM Signals — a working framework

Everyone sells you more signals. Nobody tells you which to act on.

The hard part isn’t collecting signals — it’s acting on them: which ones matter most, who to work first, and what to do when they fire. That’s what this framework is for.

You’re sorting them wrong

Classify signals by where the buyer is.

situation → behavior

Most teams sort signals by where the data came from — first-party (your site and product), second-party (partners, review sites), third-party (intent, funding, hiring feeds) — none of which tells a rep what to do.

Sort by where the buyer is on their journey instead, and one read sets the priority and picks the play.

The journey

From your hunch to their handshake.

The six stages are one buyer warming up. Quiet is a company that fits but is doing nothing — the resting state, not a signal. Pre-intent is the first flicker: a funding round or a new RevOps hire lets you guess a need is forming before anyone’s lifted a finger. From there a person has to actually act — feeling the pain, researching the category, comparing you against rivals, running diligence. Below pre-intent you’re inferring; above it, someone’s shown you.

Buyers don’t climb in a straight line — they stall, jump ahead, and slip back. So a stage is just a snapshot of where someone is today, not a permanent label — expect it to change as they move.

The GTM Signals radar — a six-stage journey WIN QUIET ≈95% · the resting state QUIET · ≈95%, RESTING PRE-INTENT · SITUATION situation → behavior line PROBLEM-AWARE · BEHAVIOR SOLUTION-EXPLORING EVALUATING DECIDING · CLOSE ON YOUR SURFACE OFF SURFACE
How to read it
  • Rings are distance along the journey — the center is a win.
  • The bold dashed circle is the situation-vs-behavior line: outside it you infer, inside it someone acted.
  • Filled dots happened on your own turf — your site, product, a demo request. Hollow ones happened out in the world — a G2 review, a funding round, a hire. It’s just where the signal occurred.
  • Accounts climb the stages over time. What matters is the direction they’re moving — one steadily working its way inward is what you’re looking for.

The six stages, in full

Each stage with an example signal, whether it’s a situation you infer or a behavior someone performed, and the score it carries.

StageWhat it meansExample signalEvidenceScore
QuietResting state — no live signal. ~95% of the market.the absence of a signal, not a signalnone
Pre-intentThe company’s situation changed; conditions may be forming.funding round · RevOps job posting · champion moved · tech-stack changesituation — you infer1
The line that matters — above: situation (a company fact you infer from) · below: behavior (a person actually acted)
Problem-awareBehavior shows they feel the pain.engaging with problem-space content, not shopping yetbehavior1
Solution-exploringBehavior shows they’re researching the category.downloaded a category guidebehavior2
EvaluatingBehavior shows they’re comparing vendors, including you.visited your pricing pagebehavior2
DecidingBehavior shows diligence or close.asked for your SOC 2 docsbehavior3

Score by stage: 1 early · 2 middle · 3 late — dock a point if the signal is ambiguous.

Why this wayYou already know the three-stage version — Awareness → Consideration → Decision. Three stages can’t separate someone researching the category from someone comparing vendors, and those two get completely different outreach. Splitting pre-intent out from problem-aware fixes a quieter mistake: a job posting isn’t proof that anyone feels the pain — it’s a company fact we’re reading intent into. Keep inference and demonstration on separate stages and the model stays honest about what each signal actually proves.
See every signal mapped in the library →

Why this matters to a revenue leader

95% of your market is quiet right now

Most of your buyers aren’t in the market right now — call it 95%, give or take by category; it’s a rule of thumb, not a law. Your competitors’ tooling — and probably yours — points almost entirely at the thin slice already showing behavior, where everyone crowds the same accounts at the same moment.

Quiet and Pre-intent are the other 95% — where the tooling is weakest and the competition thinnest. Pre-intent signals let you show up first and frame the evaluation before there’s a bake-off to lose. You can’t score your way into a quiet account. You can be the one already in the room when its situation changes.

How it works

Seven moves, and you’re scoring

It’s simpler than it looks. Classify each signal once, by its stage — that one label sets the score and picks the play. The score says whether to act; the stage says what to do.

1

Pick up to 5 signals

Start small — ones you can already detect.

And where the next move fits in a single sentence.

2

Score each signal 1–3 by its stage

Early = 1, middle = 2, late = 3.

Pre-intent and problem-aware = 1; solution-exploring and evaluating = 2; deciding = 3. If a signal could easily mean something other than buying — that pricing visit might be a competitor or a job seeker — knock it down a point. Stuck between two scores? Take the lower one and move on.

3

Tag it on-surface or off

Did it happen on your properties, or out in the world?

Set it once per signal type: on your surface (your site, product, a demo request) versus off (G2, a funding round, a hire). It doesn’t change the score — just the outreach. On-surface signals earn a warmer, more personal touch; off-surface ones a colder, more careful opener.

4

Give each signal an expiry

Start everything at 30 days; adjust as you learn.

Shorten the fast ones (a pricing visit is cold in a week), lengthen the slow ones (a funding round has a one-to-two-quarter fuse). When a signal expires it stops counting toward the score — but the record stays. Adjusting after the fact is expected.

5

Add up the live signals

The total is the account’s intent level — high, medium, or low.

Name the levels by intent rather than by action — it keeps what you measure separate from how you respond.

6

Run the play that matches the stage

Use the account’s furthest stage — how far along, not most recent.

Early gets education, evaluating gets comparison content, deciding gets a human reaching out today. Late-stage signals can jump the queue no matter what the score says.

7

Keep every signal, even expired ones

The score is today’s snapshot; the log is the memory.

It’s what lets a rep see “went quiet after asking for security docs,” and what lets you check later which signals actually ran ahead of wins.

Known limits — we accept these on purpose

Expiry is all-or-nothing. A signal counts at full value until it expires, then drops straight to zero — so an account sitting right at a band threshold can flip in and out day to day. Gradual decay would smooth it, but that’s a later refinement, not worth the complexity at low volume.

The furthest stage can over-state things. A pricing-page visitor often isn’t really evaluating yet — but sending comparison content to someone merely curious is a cheaper mistake than missing a real evaluator. A confidence score would sharpen this later.

Getting started

Crawl, walk, run

You don’t need scoring to start. One signal with a play beats five signals, a scoring model, and nobody to run them.

● Crawl

Start here

  1. Pick up to 5 signals you can already detect.
  2. Write the play for each — one sentence: what fires, and who does it.
  3. Name an owner. Signals with no owner become a dashboard nobody opens.
  4. Start acting. Put the signals where a person will actually work them — not just see them. A trickle (sales-led)? A Slack channel is plenty. Real volume (high-signal PLG)? Slack gets noisy and ignored, so a list or saved view worked like a todo queue beats it.
  5. Track whether it’s working — are signal-informed plays converting better than the same motion without them? Not just outbound you sent because of a signal, but inbound you prioritized or pitched differently because of one. The only metric that matters yet.
●● Walk

Add scoring

  1. Exclude before you score. Filter out customers, open deals, churned and non-fit accounts, internal traffic.
  2. Score and band the signals using the model above.
  3. Calibrate bands to capacity. If 400 accounts come out “high,” nobody works them.
  4. Set a repeat rule. A repeat refreshes expiry but adds no points.
  5. Pause your sequencer when a play fires, so it doesn’t step on a personal touch.
  6. Review on a rhythm — track weekly (are the plays getting worked and converting?), tune monthly, and recalibrate scores against closed-won and closed-lost quarterly, once you have a quarter of closed deals. Drop signals that show up equally before wins and losses.
●●● Run

Only when the basics work

Each one adds value. Each one adds complexity. Deferred on purpose.

  1. Confidence as its own axis — a late-stage signal can still be noisy.
  2. Frequency & velocity — three pricing visits this week beat one, and visits speeding up week-over-week beat a steady drip. Expiry already drops the stale ones; this scores intensity on top.
  3. Stacking bonuses — different signal types, or multiple people, together.
  4. PLG as its own track — self-serve product users aren’t closed customers, open deals, or cold prospects, so they slip through the exclusion filters and get mishandled. Product-usage signals need their own scoring, plays, and owner.
  5. AI-written analysis per account — an AI reads each account’s signals and writes a plain-language “here’s what’s happening, here’s the move” summary. Useful, but a layer on top — get the base model working first.
Known limitIf your high-intent actions are mostly late-stage signals jumping the queue rather than the score itself, the problem is your bands, not your signals — they’re set too loose.

Coverage audit

Which stages can you actually see?

Map your own tools onto these rows. If they cluster in one or two, whole stages of the journey are invisible to you — and you won’t realize it, because signal tools quietly cluster by stage. (A guide, not a rule: the same signal can land elsewhere depending on context.)

SourcePre-intentProblem-awareSolution-exploringEvaluatingDecidingVolumePrecision
Third-party (hiring, funding, tech, job changes)HighLow
Community & socialMediumMedium
Website visitorsMediumMedium
Product usage (PLG)HighHigh
Calls & meetingsLowVery high

● primary  ·  ○ partial  ·  — none

Reference · worked examples

Six accounts, as they’d appear on a CRM record

Signal set: hiring a RevOps leader (pre-intent, 1) · past champion joined (pre-intent, 1) · pricing page visit (evaluating, 2) · demo attended (evaluating, 2) · security docs requested (deciding, 3). The later stage always sets the play.

Acme Corp

EvaluatingScore 3 · Medium
  • Hiring RevOps lead16d ago
    Pre-intent · 1Off surface
  • Pricing page visit2d ago
    Evaluating · 2On your surface

Evaluating-stage nurture — comparison content. The later stage wins, so skip the pain-awareness material; and since the pricing visit is on-surface, warm the channel up.

Beacon Labs

EvaluatingScore 5 · Medium
  • Past champion joined (VP Ops)12d ago
    Pre-intent · 1Off surface
  • Pricing page visit2d ago
    Evaluating · 2On your surface
  • Demo booked (colleague)1d ago
    Evaluating · 2On your surface

Warm personal outreach — founder or AE reaches out personally and leans on the old relationship. No cold sequence.

Champion-moved is pre-intent, so it scores 5, not 6 — a stacking bonus would bump it later.

Nimbus Inc

DecidingScore 5 · jumped queue
  • Requested SOC 2 docs1d ago
    Deciding · 3On your surface
  • Pricing page visit7d ago
    Evaluating · 2On your surface

Human touch, same day — send the docs, offer a procurement call. The deciding-stage signal trumps the band.

Orbita

QuietScore 0 · Low
  • Posted sales ops role45d ago
    Pre-intent · 0Off surface
  • Pricing page visit38d ago
    Evaluating · 0On your surface

No play — they dropped out on their own when the signals expired. The history stays: evaluating back in March, handy context the moment something new fires.

Vertex

EvaluatingScore 4 · Medium
  • Pricing visits (3 people, counts once)5d ago
    Evaluating · 2On your surface
  • Demo attended3d ago
    Evaluating · 2On your surface

Evaluating nurture — use the multi-person activity as your angle: “your team has been evaluating.”

Stacking bonuses would lift this one.

Talos Systems

Solution-exploringScore 1 · Low
  • Downloaded "scaling outbound" guide (docked — ambiguous)4d ago
    Solution-exploring · 1On your surface

No queue entry — passive category education, not comparison content; they’re still researching the category, not weighing vendors.

A three-stage model would send the wrong thing.

Bands are illustrative — yours come from calibrating against capacity: 0–2 low · 3–5 medium · 6+ high.

The library

117 signals, already mapped.

You’ve got the framework. The library is where it pays off — a working catalog with every signal placed on the journey and paired with the play to run. Open it instead of starting from a blank page.

Every signal carries
  • Stage
  • Play to run
  • Decay speed
  • Noise level
  • Effort to act
  • On / off surface
Browse the signal library →