GTM Engineering · Guide

Most buying signals are noise. Eight are worth acting on, and every one has a half life

A signal is only worth a send while it is still true. This is the list we actually watch, how each one is detected, and how fast each one decays.

By Ishan Vats · Founder of IV Consulting · builds signal based outbound systems

Updated Oct 2026 11 min read Pillar: GTM Engineering

See how we map your signals → Jump to the list

8 signals that hold up Half life per signal What we run in Clay What we deliberately cut

Signal decay · the window you have

Hour 0Signal fires, publicly

Days laterA weekly batch notices

72 hrsSite visit, gone

30 daysStack change, cooling

90 daysJob change, live

Same signaldifferent shelf life

Quick answer

The buying signals worth acting on are the ones you can write a specific first line from, and there are far fewer of them than most teams think. Eight hold up in practice: a decision maker changing role, a funding round, a fund closing a new vehicle, hiring for the function you serve, a tech stack change, repeat visits to your pricing or docs, a former champion landing somewhere new, and a leadership departure in the team you sell to. What separates these from the hundred other things you could track is not accuracy. It is half life. A pricing page visit is worth acting on for roughly seventy two hours. A funding round is worth acting on for a quarter. Track a signal without knowing how fast it decays and you will send into a window that has already closed, which is the most common reason signal led outbound underperforms the pitch. In our own Clay systems, the signal doing the most work is a job post that names the exact thing we sell.

The list

Which buying signals are actually worth acting on?

Every row here passes one test: you can write a first line from it that could only be sent to that account, this week. Anything you can only reference in passing did not make the list.

Signal	Why it converts	How it is detected	Half life we work to
Decision maker changes role	New owner, new priorities, and a budget that resets with them.	LinkedIn and enrichment provider change feeds	About 90 days, strongest in the first 30
Funding round announced	New capital means a spending plan being written rather than executed.	Funding databases and news feeds via API	One quarter
A fund closes a new vehicle	The money is new and the thesis is public before deployment starts.	Regulatory filings and fund announcements	One quarter, strongest in the first three weeks
Hiring for the function you serve	An open role is a public admission of the gap your product fills.	Careers pages and job boards, scraped where there is no API	While the role is open
Tech stack change	A tool added or dropped next to yours changes what is possible and what hurts.	Technographic lookups and site fingerprinting	About 30 days
Repeat visits to pricing or docs	Somebody is evaluating you right now, without telling you.	Your own analytics and de-anonymisation, wired to the pipeline	About 72 hours
Former champion in a new role	Someone who already bought from you, now holding a different budget.	Your CRM matched against change feeds	About 90 days
Leadership departure in the function	The team reassesses vendors while the seat is empty, not after it is filled.	Change feeds plus press	About 60 days

Be careful with that last column Those half lives are our own operating windows, set from running these systems, not published benchmarks. Treat them as a starting cadence to measure against rather than a finding to quote. Yours will differ by market, and the point is that you should know what yours are.

Our verdict

Which buying signals do we actually run in Clay?

IV runs six buying signals across three live Clay systems: an open content role, a blog gone quiet, a competitor shipping an MCP server, a job post naming MCP, a dated company update inside 90 days, and a fund closing a new vehicle. The table above is the long list. This is the short one, as of October 2026, with how each signal is weighted and what we cut.

System	Signal it acts on	How Clay detects it	How we weight it
Blog cadence (25 companies a day)	An open content writer or copywriter role	A signal first search lane for US and UK companies with 10 to 99 staff, recruiters and agencies excluded	+3 once open 30+ days. A filled role plus a welcome post inside 60 days disqualifies the company
Blog cadence	A company blog that went quiet	An AI agent reads the recent post dates	+3 went quiet, +1 always sparse. Supporting: ad pixel +2, LinkedIn active while the blog is dead +2, SEO tool +1
MCP prospecting (10 companies a day)	A direct competitor shipped an MCP server	Two research gates check the company and its 2 closest peers against their own docs, GitHub, npm and the official MCP registry	The hook itself. No verified competitor MCP means no email, and the company is rechecked in 60 days
MCP prospecting	A job post that names MCP	A dedicated daily lane of 3 companies	+15, and it picks the technical buyer and the angle. A generic DevRel or API role counts for nothing on its own
MCP prospecting	The company's own dated update in the last 90 days	An AI research step, with a code guard that throws out undated facts and forecasts	+30, the single largest weight in the score
VC partner	A fund closing a new vehicle	A scheduled watch on fund announcements, enriched to the partner covering the sector	The whole system. A guard checks the partner's email domain matches the firm

The verdict If you can only build one detector, build the one for a job post that names the exact thing you sell. It is public, cheap to detect, its half life is visible (the role is open or it is not), and it tells you who to write to. Second, a dated change at the account inside a window you enforce in code, like our 90 day rule. Third, peer pressure: a named competitor who already did the thing. The signal we would cut first is LinkedIn engagement. We planned it as the top signal in the MCP system and dropped it before launch, because Clay cannot read it without scraping and we do not automate LinkedIn.

What building them taught us

Score, do not gate. Every supporting signal adds points, and no signal is neutral rather than a strike. The only hard gates are the ones that mean the problem is already solved: a content hire, or an MCP server of their own. The point values are starting guesses, and the plan is to recalibrate them once 30 replies are in.

Re-check state signals at send time. A quiet blog is a state, not an event, and states expire without telling you. When we audited our own four Clay systems on 13 October 2026, 26 of the 71 companies in the quiet blog hot tier had posted in September or October by the time the email went out. The signal was true when it was scored and false when it was sent.

Where you look matters as much as what you look for. In a 45 company lane test on the MCP system, 7 of the 15 companies from a broad AI SaaS sweep already had an MCP server, while the vertical and ops SaaS lane had none and the best qualified rate, 10 of 15. That is a small, first pass sample, but it was enough to move the daily mix to 6 vertical companies a day.

A wrong fact costs more than a missed lead. Leads with an undated or forecast growth fact, a hook on a category giant rather than a direct peer, or no verified email sit on hold and never enter a campaign. Enrichment only runs on the accounts that clear the gates, which is also what keeps cost per verified contact under control.

What we cannot tell you yet Which of these signals predicts a reply. At the 13 October audit the systems had sent 471 emails to 272 people with 0 positive replies, one to two weeks in. We publish that on purpose: the weights above are operating decisions, not proof, and this page will be updated when the reply data says which ones to move.

Before you copy our list Your signals should come from your buyers, not from ours. See how we map the buying signals for your ICP, then build the detection, scoring and CRM writeback in your own Clay and CRM accounts.

See the GTM engineering setup →

The mechanics

What is signal half life, and why does it decide your cadence?

Half life is how long a signal stays true enough to be worth writing about. It is the difference between a message that reads as attentive and the same message, three weeks later, reading as automated.

Most teams treat detection as the hard part and then send on a weekly schedule that has nothing to do with the signal they detected. That is backwards. Detection is largely a solved problem you can buy. Cadence is the part you have to decide, and the decay rate is what should decide it. Wiring that decision into detection, scoring and sending is the core of what GTM engineering actually is.

Fast decay signals need a daily run

A repeat visit to your pricing page is worth about seventy two hours. Somebody is comparing you against two other options in a tab right now. Detect that on Monday, send on Friday, and you are writing to a decision that has already been made. Anything with a window measured in hours or days has to run daily, which means it has to run without a person remembering to run it.

Slow decay signals can run weekly

A decision maker who changed role three weeks ago is still new. Their budget is still forming. A weekly batch is fine, and a daily one wastes sending capacity you would rather spend on the fast signals. Splitting your cadence by decay rate is usually the single cheapest improvement available, because it costs nothing except deciding to do it.

IV Consulting take The number that exposes this is time to first touch. Measure the gap between a signal firing and your first message leaving, per signal type. In most estates we look at, that gap is measured in days while the signal is measured in hours, and nobody has ever put the two numbers next to each other. That comparison is usually the first slide of our diagnostic, and it is the reason our GTM and CRM setup starts with a signal map rather than a list.

The cut list

What looks like a buying signal but is not one?

The first thing we do on a build is cut the signal list, not extend it. Six triggers you can write a specific opening line from will beat sixty you can only mention.

Company size and industry

That is targeting, not timing. It tells you who to contact and says nothing about when. Useful in a filter, useless as a trigger.

A generic content download

Somebody wanted a PDF. Treating that as intent is how a prospect gets a five email sequence for reading one page, and how your domain earns a reputation you cannot undo quickly.

A LinkedIn post like

You cannot write a first line from it that does not sound like surveillance. If referencing the signal makes the recipient uncomfortable, it is not a signal, it is a trace.

Anything already in every sequence

If three competitors detect the same trigger from the same vendor and send in the same week, you have bought a crowd, not an advantage.

There is one test that settles almost all of these. Write the opening sentence the signal would produce. If that sentence could be sent to any other account on your list without changing a word, the signal is not doing any work and you are paying for it twice: once to the data vendor and once in reputation.

The same restraint applies to the channel. A strong signal earns a short note a human sends from their own LinkedIn profile, not a place in an automation tool's queue. That is the line we draw in HeyReach vs Expandi vs manual outreach after the LinkedIn ban.

Worked example

What does one signal look like end to end?

Take the third row of the table, a fund closing a new vehicle, because it is a signal almost nobody writes into and the timing is unusually clean.

Venture funds announce new vehicles in public, in regulatory filings and in the press. Between that announcement and the fund actively writing cheques there is a short window when the capital is new, the thesis is stated, and the partner covering your sector has a mandate they have not spent yet. Write in that window and the message is obvious. Write a quarter later and the allocation is already committed.

We ran exactly this for a founder raising capital. The signal was watched on a schedule so a close was picked up in days rather than found later by accident. Each one was enriched to the individual partner covering that sector and stage, not a general inbox. The first touch went out while the money was new.

35meetings booked with positive intent, inside two weeks

$850kraised

1signal, watched properly, rather than eight watched loosely

The lesson is not that fund closes are a magic trigger. It is that one signal with a known window, detected on a schedule and enriched to a named person, outperforms a long list of signals nobody has assigned a cadence to. That is the whole argument of building outbound as a system rather than as a campaign. It also keeps the data bill small: you only pay to enrich the accounts that fired, which matters once you see what one verified contact costs in Clay versus FullEnrich or n8n.

The plumbing

How do you detect buying signals without buying eight tools?

Three sources cover everything in the table, and most teams already pay for one of them.

1. Providers with an API

Job changes, funding rounds and technographics are sold by data vendors and arrive clean. Pull them on a schedule into one table rather than reading them in a vendor dashboard, because a signal you have to log in to see is a signal you will act on late.

2. Scraping, for the things nobody sells

Careers pages, directories, filings and listings are public and usually have no API. This is where a scraper earns its place, and it is often the source of the signals your competitors do not have, precisely because they are not for sale.

3. Your own first party data

Repeat visits to pricing and docs, and your CRM's history of people who already bought from you, are the two highest intent sources on the list and you already own both. The work is wiring them into the same pipeline as everything else instead of leaving them in an analytics tab. If your CRM is not built to receive that, start with the CRM: we cover what that setup involves in GTM and CRM setup, and the glue between tools you already pay for in automation.

One table, not eight dashboards Whatever the sources, land them in a single place with a detection date on every row. The date is what makes half life measurable, and without it you cannot tell the difference between a signal that did not work and a signal you sent to three weeks late.

FAQ

Questions people ask about buying signals

What is a buying signal in B2B sales?+

A buying signal is a specific, observable event at an account that gives you a defensible reason to contact them this week rather than any other week. A decision maker changing role, a funding round closing, or a repeat visit to your pricing page all qualify. Company size and industry do not, because they tell you who to contact and nothing about when.

How many buying signals should we track?+

Fewer than you want to. Our MCP prospecting system scores six inputs and our blog cadence system scores five, and in each one a single signal does most of the work. Six triggers that produce a genuine reason to write this week will outperform sixty you can only reference in passing. The constraint is not detection, it is that every signal you add needs its own cadence, its own copy and its own measurement, and teams that track thirty end up acting on four of them properly.

What is signal half life?+

Half life is how long a signal stays true enough to write about. A repeat pricing page visit is worth acting on for roughly seventy two hours. A funding round is worth acting on for a quarter. The decay rate should set the cadence for that signal, which means fast decaying signals have to run daily and without a person remembering to run them.

Do buying signals actually improve results?+

We will not quote you an industry average, because the ones in circulation are measured differently by every vendor publishing them. What we can point at is our own: 35 meetings booked with positive intent inside two weeks, and $850,000 raised, from one signal watched properly rather than eight watched loosely. Judge the method by the mechanism rather than by a benchmark.

Can we track buying signals without a data vendor?+

Partly. Your own analytics and your CRM cover the two highest intent signals on the list, repeat visits to pricing and former champions in new roles, and you already own both. Careers pages and filings can be scraped. Job change and funding data is where a provider genuinely earns its fee, because maintaining that yourself costs more than it costs to buy.

What is the first thing to fix if our outbound is underperforming?+

Measure time to first touch per signal type: the gap between a signal firing and your first message leaving. In most estates we look at, that gap is measured in days while the signal is measured in hours. Fixing the cadence usually beats adding signals, and it costs nothing except deciding to do it.

Who wrote this

Ishan Vats

Founder, IV Consulting · AI & automation consultant

I build the GTM and AI systems teams run on, from signal detection through to CRM writeback. 150+ engagements over 10+ years. Every half life in the table above comes from running these systems, not from a vendor deck.

See how we build it →

Keep reading

Related guides and work

AI and automation use cases for sales and marketing teams

Guide

AI and automation for sales and marketing teams

Where the automation actually pays back once the signals are landing.

Read the guide →

Stack

Scraping the signals nobody sells you

The build for the sources that have no API and never will.

See the stack →

Responding to inbound leads automatically

Build

Answering a fast signal in minutes, not days

What it takes to respond while the window is still open.

Read the build →

Want to know which signals you are already missing?

Book the signal audit, a free 30 minute call. We will look at who you sell to, name the signals firing on those accounts that nobody at your company is watching, and tell you what the window is on each one. If the honest answer is that your market does not produce enough signal to build on, we will say that instead.

Book the signal audit →

Free 30 minute call. You leave with the signal list either way.