How to Build Custom Buying Signals for B2B Outbound (Not Generic Intent Data)

Most B2B Outbound Teams Are Running on the Same Signals. That Is the Problem.

Every outbound team on the market has access to funding data. Every SDR tool tracks hiring spikes. Every intent platform monitors technology installs and executive changes. These signals are everywhere, which is exactly why they have stopped working as a competitive advantage.

When a company raises a Series B, fifteen vendors email the CRO the same week. "Congrats on the raise, here is how to spend it." That is not signal-led outbound. That is slightly-less-cold outbound with a better subject line.

The problem is not that generic signals are wrong. Funding rounds, hiring patterns, and tech changes are real events. The problem is that they are commodity data. If your signal model is built from the same vendor databases every other team uses, you are competing on send speed alone. Somebody will always be faster.

We have built signal-led outbound systems for over 30 B2B companies at AdvancedClient. The pattern is consistent: custom buying signals built from a client's own sales data outperform generic intent data by 3-5x on reply rates and meeting conversion. Not because the tooling is better. Because the signals are specific to what actually predicts a deal closing for that particular business.

This article covers how to build those custom signals from scratch. The full process, worked examples, the tools, the scoring framework, and the mistakes that break signal models in production.

A custom buying signal is a weighted indicator, derived from analysis of a company's own closed-won deals, closed-lost deals, and sales call transcripts, that identifies the specific conditions under which a target account is likely to purchase.

What Generic Intent Data Actually Tells You (and What It Does Not)

Generic buying signals are events you can track using off-the-shelf tools. They include:

• Funding rounds (Crunchbase, PitchBook, Clay)
• Hiring activity (LinkedIn, Indeed, job board scrapers)
• Technology changes (BuiltWith, Wappalyzer, technographic providers)
• Executive moves (LinkedIn, press releases)
• Website visits (RB2B, Vector, Clearbit Reveal)
• Content engagement (G2, Capterra, review site activity)

These are useful as a starting layer. They tell you a company is changing. They do not tell you a company is ready to buy what you sell.

Consider what a funding round actually signals. A SaaS company raising a Series B does not mean they need outbound infrastructure. They might invest in product. They might hire engineers. They might already have a system that works. Funding is a financial event, not a buying signal for your specific category.

The same logic applies to hiring. A company posting one Customer Success role is a different signal from a company posting three SDR roles and a VP Sales. The first indicates retention investment. The second indicates outbound buildout. But even the second is only relevant if you sell something connected to outbound infrastructure.

Generic signals answer the question "is this company changing?" Custom signals answer the question "is this company likely to buy from us right now?" Those are fundamentally different questions, and the second one is the only one that matters for outbound performance.

Why Generic Signals Have Degraded Over Time

Three years ago, tracking funding rounds as a buying signal gave you a genuine edge. Most outbound teams were still running static lists. Today, every modern SDR tool includes signal tracking as a default feature. Clay, Apollo, Trigify, ZoomInfo, and dozens of others all surface the same underlying data.

The result is signal saturation. A prospect who matches common signal criteria (new funding, new executive, hiring spike) receives dramatically more outbound than a prospect who does not. The signal that used to indicate buying intent now mostly indicates that the prospect's inbox is full of pitches.

This is the core argument for building custom signals. Not to replace generic signals entirely, but to find the indicators that are specific to your market and invisible to your competitors.

What a Custom Buying Signal Actually Looks Like

A custom buying signal is not "they raised funding." It is a specific combination of conditions that you have identified, through analysis of your own sales data, as predictive of a deal closing.

Here are three real examples of what custom signals look like in practice, drawn from systems we have built:

Example 1: Usage-based pricing platform Generic signal: "Company raised Series B." Custom signal: "Company is replacing their current billing system + hired a pricing analyst in the last 60 days + running a usage-based model on a legacy billing stack."

The generic signal tells you they have money. The custom signal tells you they have a specific operational problem that this product solves. The outreach for the custom signal references their billing stack migration. The outreach for the generic signal says "congrats on the raise."

Example 2: Sales training firm (RAIN Group) Generic signal: "New VP Sales started." Custom signal: "Company expanded sales team by 30%+ in 6 months + sales cycle length increased quarter-over-quarter + existing sales methodology is outdated or absent (identified from job posting language and Glassdoor reviews)."

This was one of the buying conditions we modelled during a 12-week engagement with RAIN Group. The generic signal (new VP Sales) triggers outreach from every training vendor. The custom signal identifies the specific conditions where RAIN Group's training programmes had historically produced the highest ROI.

Example 3: Background screening company (Verifile) Generic signal: "Company is hiring." Custom signal: "Company operating in a regulated industry + headcount growth exceeding 20% in 90 days + no existing enterprise screening provider detected in their tech stack + compliance-related job posting language present."

The generic signal applies to thousands of companies. The custom signal narrows to companies where Verilife's specific product and compliance expertise is relevant. The result with Verifile: $312K in pipeline generated within 28 days.

The Question Every Prospect Asks

When we present signal-led outbound to prospective clients, the most common question is: "What does a signal look like in practice?"

The answer is always: it depends on your business. That is not a dodge. It is the entire point. A buying signal for a sales training firm looks nothing like a buying signal for a usage-based billing platform. The signals that predict deals for your company are specific to your market, your product, your buyers, and your sales process. They cannot be bought off the shelf because they do not exist until someone builds them from your data.

Generic Signals vs Custom Signals: The Full Comparison

Dimension	Generic Signals	Custom Signals
Source	Third-party vendor databases (Crunchbase, LinkedIn, BuiltWith)	Your own CRM, call recordings, closed-deal analysis
Specificity	Category-level ("they are changing")	Company-level ("they match the pattern of our best deals")
Competitor access	Every team using the same vendors sees the same signals	Unique to your business. Competitors cannot replicate.
Signal volume	High. Thousands of accounts trigger generic signals daily.	Lower. Fewer accounts match, but those that do convert at much higher rates.
Outreach relevance	"Congrats on the raise. Here is what we do."	"You are replacing your billing stack and hired a pricing analyst. Here is how we help companies in that exact situation."
Reply rate (observed)	2-4% on cold outbound triggered by generic signals	8-15% on cold outbound triggered by custom signal matches
Meeting conversion	15-25% of replies convert to meetings	35-50% of replies convert to meetings
Time to build	Minutes (turn on a vendor feed)	4-12 weeks (requires sales data analysis and modelling)
Maintenance	Low. Vendor manages the data.	Medium. Model needs quarterly review and recalibration.
Best use	First-pass filtering. Identifying accounts that are changing.	Second-pass scoring. Identifying accounts that are likely to buy from you specifically.

The performance difference is not marginal. Custom signals consistently produce 3-5x the results of generic signals across the 30+ systems we have built. The trade-off is time: building a custom signal model takes weeks, not minutes. But the pipeline it generates is fundamentally different in quality.

The 4-Step Process to Build Custom Buying Signals

Building a custom signal model is a research project, not a software configuration. It requires access to sales data, time to analyse it, and the discipline to let the data tell you what matters rather than guessing.

Here is the process we use at AdvancedClient for every signal build.

Step 1: Analyse 50+ Closed-Won Deals

Pull every deal your team closed in the last 12-18 months. You need at least 50 to identify reliable patterns. Fewer than 50 and you are working with anecdotes, not data.

For each closed-won deal, document:

• Company characteristics at the time of first contact: Industry, revenue range, headcount, growth rate, funding stage, geography
• Operational conditions: What was happening inside the company when they bought? Were they replacing an existing tool? Building a new team? Responding to a regulatory change? Scaling after a funding round?
• Trigger event: What specifically prompted them to evaluate your product? Was it internal (new leader, broken process) or external (market shift, competitor move)?
• Buying committee: Who was involved in the decision? What were their titles? Who championed the deal internally?
• Sales cycle length and dynamics: How long from first touch to close? What objections came up? What accelerated the deal?
• Technology context: What tools were they using? What were they replacing or adding to?

Once you have this for 50+ deals, you start looking for patterns. Not the obvious ones (they were in the right industry, they were the right size). The specific conditions that appear repeatedly across your wins.

What you are looking for:

The conditions that appear in 60%+ of your closed-won deals are your candidate signals. If 35 of your 50 wins involved a company that had recently expanded their sales team by more than 20%, that is a signal. If 40 of 50 wins involved a company replacing an existing vendor, that is a signal. If 28 of 50 wins had a new CRO who started within the previous 6 months, that is a signal.

Write down every condition that appears in more than half your wins. You will have 15-25 candidate signals at this stage.

Step 2: Analyse Closed-Lost Deals

Now pull every deal your team lost in the same period. Same documentation framework.

The goal here is different. You are looking for what your lost deals were missing compared to your wins. This is where the signal model gets its negative filters.

Common patterns in closed-lost analysis:

• The company was too early in their buying journey (exploring, not ready to act)
• A key internal condition was missing (no budget owner, no executive sponsor, wrong team structure)
• They were evaluating for a different use case than your product's core strength
• The operational trigger was absent (they were "interested" but nothing was forcing a decision)
• Timing was wrong (they had just signed a competing contract, or were mid-migration on something else)

What you are looking for:

The conditions that appear in your losses but not your wins become exclusion criteria in your signal model. If 70% of your lost deals involved companies with fewer than 50 employees, that is a negative signal. If 80% of your losses involved companies that had recently signed a new contract in your category, that is a negative signal.

This step is where most teams skip or rush. Do not skip it. The exclusion criteria are as valuable as the positive signals because they prevent you from wasting outbound on accounts that look good on paper but will not close.

Step 3: Mine Sales Call Transcripts for Buying Language Patterns

This is the step that most signal-building guides leave out entirely, and it is the step that produces the highest-value signals.

Pull transcripts from at least 100 sales calls. Split them into two groups: calls that led to closed-won deals and calls that led to closed-lost deals or stalled opportunities.

In the closed-won transcripts, look for:

• Problem language: How did buyers describe their pain before they bought? What words and phrases did they use? "We are outgrowing our current process." "Our team is spending 40% of their time on manual research." "We tried three tools and none of them worked."
• Urgency indicators: What phrases indicated they were ready to act now, not later? "We need this sorted before Q4." "Our board is asking about pipeline." "We have budget allocated."
• Decision triggers: What specific events or realisations pushed them from evaluation to purchase? "Our CRO said we need to fix this by end of quarter." "We lost a deal because of this."
• Objections that resolved: What concerns came up that were successfully addressed? These tell you what a qualified buyer worries about versus an unqualified one.

In the closed-lost transcripts, look for:

• Exploratory language: "We are just looking at options." "This is more of a long-term project." "We do not have a timeline yet."
• Misalignment indicators: "We were hoping you could also do X." "Our main problem is actually Y." These indicate the prospect's problem did not match your product's strength.
• Stall language: "Let us revisit this next quarter." "We need to get more stakeholders involved." "Can you send me some resources to share internally?"

How to turn transcript analysis into signals:

The buying language patterns become inbound signal indicators. When a prospect uses the same phrases your closed-won buyers used, they are further along the buying journey than a prospect who uses exploratory language. You can monitor for these patterns in discovery calls, in LinkedIn posts, in webinar Q&A, and in support tickets.

Example from a system we built: prospects who said "we are spending X hours per week on manual research" in their first call closed at 3.2x the rate of prospects who said "we are interested in automating some of our processes." Same product, same price point, same team selling it. The language pattern was a signal.

Step 4: Model the Signals and Weight Them by Conversion Correlation

You now have three data sets: closed-won patterns, closed-lost exclusion criteria, and buying language indicators. The final step is combining them into a weighted model.

How to weight signals:

Not all signals are equal. A signal that appeared in 90% of your closed-won deals is a stronger predictor than one that appeared in 55%. A signal that appeared in 0% of your closed-lost deals is an even stronger indicator (it perfectly separates wins from losses).

For each candidate signal, calculate:

Signal	% Closed-Won	% Closed-Lost	Conversion Lift	Weight
Replacing existing vendor in your category	78%	12%	6.5x	High (3 points)
New CRO started within 6 months	62%	18%	3.4x	High (3 points)
Sales team expanded 20%+ in 90 days	58%	22%	2.6x	Medium (2 points)
Active in relevant industry communities	44%	30%	1.5x	Low (1 point)
Recently attended competitor event	36%	34%	1.1x	Drop (not predictive)

Conversion lift is the ratio of presence in closed-won vs closed-lost. A signal that appears in 78% of wins and 12% of losses has a 6.5x lift, meaning accounts with that signal close at 6.5x the rate of accounts without it.

Weight assignment:

• 3x+ conversion lift = High weight (3 points)
• 2-3x conversion lift = Medium weight (2 points)
• 1.5-2x conversion lift = Low weight (1 point)
• Below 1.5x conversion lift = Drop it. It is not predictive enough to justify tracking.

The resulting model gives every target account a composite score based on which signals they match and how heavily weighted those signals are.

What a Custom Signal Model Looks Like in Practice

The best way to explain the output is through a real build.

The RAIN Group System

RAIN Group is a global sales training firm. When they engaged AdvancedClient, they had a mature sales organisation but were running traditional outbound: static lists, manual research, and periodic email campaigns.

Over a 12-week engagement, we built a signal-led outbound system with the following specifications:

• 15 live buying signals tracked continuously
• 3 buying conditions modelled from RAIN Group's sales data
• 7 tools built and integrated into a unified infrastructure
• 100% ownership transferred to RAIN Group at handover

What the buying conditions looked like:

Each buying condition was a combination of multiple signals that, when they appeared together, indicated a high probability that the account would benefit from RAIN Group's training programmes. These were not generic. They were derived from analysing what RAIN Group's best clients had in common before they became clients.

One buying condition (anonymised) combined:

1. Sales team expansion above a threshold in a defined time window
2. A specific pattern in how the company described their sales challenges in public content
3. An operational indicator from their job postings that suggested their current methodology was under strain
4. A growth metric that matched the profile of RAIN Group's highest-LTV clients

No competitor could replicate this signal model by subscribing to the same data sources. The model was built from RAIN Group's own sales history. The signals were specific to their market, their buyers, and their sales process.

Jason Murray, CSO at RAIN Group, described the work as "outstanding quality, tailored to our business."

What Happens When a Signal Fires

When a buying condition is met (multiple signals fire for the same account), the system does not just add the account to a list. It routes the account to a specific play.

The play includes:

• Which sequence to trigger (tailored to the buying condition, not a generic follow-up)
• Which messaging angle to use (based on which signals fired, the outreach references the prospect's actual situation)
• Which channel to prioritise (some buying conditions route to email, others to LinkedIn, others to phone)
• Which rep should own the account (based on territory, vertical expertise, or deal size)

This is the difference between a signal model that feeds a CRM and a signal model that runs an outbound system. The signal does not just tell you "this account is interesting." It tells you what to do about it.

Tools for Building and Operationalising Custom Signals

Custom signal building requires a different toolchain than generic signal tracking. Here is what each layer does.

Signal Modelling

WhiteWhale is the tool we use to define, weight, and operationalise buying signals that are specific to a client's market and sales data. It takes the patterns you find in closed-won analysis and turns them into a working signal model. It is not a web traffic tracker or a generic intent platform. It builds custom models.

Data Enrichment

Clay is the enrichment layer. Once you have defined your signals, you need data to detect them. Clay connects to 75+ data providers and lets you build enrichment workflows that check target accounts against your signal criteria in real time. If one of your signals is "company hired a pricing analyst in the last 60 days," Clay can monitor for that specific condition.

LinkedIn Signal Detection

Trigify monitors LinkedIn for specific events: job changes, company updates, engagement patterns, and content signals. If one of your custom signals involves a prospect engaging with content about a specific topic, or a new executive posting about their priorities, Trigify surfaces that.

Web Traffic Identification

RB2B identifies which companies and individuals are visiting your website. This is a different signal category from custom buying signals, but it is useful as a reinforcement layer. When a company that matches your custom signal model also visits your website, the composite signal is very strong.

Sequencing and Routing

Instantly handles the outbound execution. When a signal fires and routes an account to a specific play, Instantly manages the email sequences, deliverability, and follow-up cadence.

How the Tools Fit Together

The architecture follows this flow:

1. WhiteWhale defines the signal model (what to look for)
2. Clay enriches target accounts against those signals (checking the data)
3. Trigify monitors for LinkedIn-based signals in real time
4. RB2B adds web visit data as a reinforcement signal
5. When signals fire, accounts route to the correct play in Instantly or through LinkedIn via manual or automated outreach

The tooling is important, but it is not the hard part. The hard part is the research that defines the signals. Getting the right tools is a configuration problem. Getting the right signals is a research problem.

How to Score Accounts Using Custom Signals

Once your signal model is built, you need a scoring framework that turns signal data into action. We use a three-tier system.

Tier 1: High-Priority (Multiple Custom Signals Firing)

Criteria: Account matches 2+ high-weight custom signals, or 1 high-weight signal + 2 medium-weight signals. Composite score above 7 points.

Action: Immediate outbound. Personalised sequence referencing the specific signals that fired. Multi-channel (email + LinkedIn + phone if the account is high-value). Senior rep assignment.

Example: A target account is replacing their existing vendor in your category (3 points) + their new CRO started 4 months ago (3 points) + sales team expanded 25% in the last quarter (2 points). Total: 8 points. This account gets a highly personalised, multi-touch sequence within 48 hours.

Tier 2: Medium-Priority (Some Custom Signals + Generic Signals)

Criteria: Account matches 1 high-weight custom signal, or 2+ medium-weight signals. Some generic signals also firing (funding, hiring). Composite score 4-6 points.

Action: Standard signal-triggered sequence. Personalised but not fully bespoke. Email-first with LinkedIn touchpoints. Regular rep rotation.

Example: A target account hired a pricing analyst (2 points) + recently attended a competitor event (not scored, but noted for messaging) + raised a Series B (1 point as a generic signal). Total: 3 points from custom signals + generic context. This account enters a tailored but templated sequence.

Tier 3: Low-Priority (Generic Signals Only)

Criteria: Account matches generic signals (funding, hiring, tech changes) but no custom signals. Composite score below 4 points.

Action: Add to monitoring. Do not trigger outbound yet. Continue enriching against custom signal criteria. These accounts may become Tier 1 or Tier 2 as conditions change.

Example: A target account raised funding and posted three new roles. Generic signals are firing, but none of your custom signals match. This account is changing, but you have no evidence they need what you sell specifically. Monitor, do not message.

Worked Scoring Example

Here is how scoring plays out for a hypothetical target account list of 500 companies:

Tier	Accounts	% of Total	Expected Reply Rate	Expected Meetings
Tier 1	35-50	7-10%	10-15%	12-20
Tier 2	80-120	16-24%	5-8%	8-15
Tier 3	330-385	66-77%	2-3%	Monitoring only

The maths is straightforward. If you message all 500 accounts with the same generic sequence, you get a blended 2-3% reply rate and maybe 8-12 meetings. If you prioritise Tier 1 accounts with personalised outreach, you generate 12-20 meetings from 35-50 accounts. Better meetings from fewer sends with less inbox burn.

Connecting Signals to Sequences

A signal model without routing rules is a dashboard. It tells you what is happening but does not tell anyone what to do about it. The final step in building a production system is connecting each signal combination to a specific outbound play.

If-Then Signal Routing

The routing logic follows a simple structure:

If [signal combination fires] then [trigger this specific play].

Examples:

Signal: Account is replacing existing vendor in your category + new CRO started within 6 months. Play: "Vendor replacement" sequence. Messaging focuses on transition risk, migration support, and what other companies in their position chose. 5-touch email sequence + LinkedIn connection request from a senior team member.

Signal: Sales team expanded 20%+ in 90 days + sales cycle length increased QoQ. Play: "Scaling pain" sequence. Messaging focuses on the operational challenges of scaling a sales team quickly without the supporting infrastructure. 4-touch email sequence + 2 LinkedIn touchpoints.

Signal: Prospect used buying language in a discovery call ("we are outgrowing our current process" or similar high-conversion phrases). Play: "Warm inbound" sequence. Accelerated timeline. Direct CTA to book a technical conversation. 3-touch sequence with a focus on speed and specificity.

Signal: Company matches multiple Tier 1 custom signals but has not responded to email. Play: "Multi-channel escalation." LinkedIn voice note from a senior team member. Direct mail if the account is high enough value. Phone call from an SDR who has been briefed on which signals fired and why this account is a priority.

Why Routing Matters More Than You Think

Most teams that build signal models stop at the scoring step. They have a ranked list of accounts. The SDRs then write personalised emails by hand, using the signal data as context.

This works at low volumes but breaks at scale. If you have 40 Tier 1 accounts firing signals every week, you cannot hand-craft 40 unique sequences. The routing layer automates the play selection so that the right message goes out at the right time without requiring manual intervention for every account.

The system we built for m3ter used this routing approach. Custom signals fired, accounts were scored, and the appropriate sequence was triggered automatically. The result: $2.4M in pipeline, 42 meetings booked, $447K closed-won.

Common Mistakes That Break Signal Models

We have built enough of these systems to know where they fail. Here are the nine most common mistakes.

Mistake 1: Using Generic Signals as Your Primary Model

This is the most common error. Teams set up funding and hiring alerts, call it a "signal model," and wonder why results are mediocre. Generic signals are a starting layer. They are not a model. If your best signal is "they raised funding," you do not have a custom model yet.

Mistake 2: Not Updating the Model

Buying patterns change. Markets shift. What predicted a deal closing 12 months ago may not predict it today. Signal models need quarterly review. Pull fresh closed-won and closed-lost data every quarter and check whether your signals still hold. If a signal's conversion lift has dropped below 1.5x, deprecate it. If new patterns have emerged, add them.

Mistake 3: Too Many Signals

More signals is not better. If you are tracking 30 signals, you have noise, not a model. The ideal range is 10-15 active signals grouped into 3-5 buying conditions. Each signal should have a clear conversion lift and a defined weight. If you cannot explain why a signal is in the model and what it predicts, remove it.

Mistake 4: Building Signals from Assumptions, Not Data

"I think companies that attend conferences are good prospects" is not a signal. It is a guess. Every signal in your model should be validated against your closed-won data. If conference attendance does not appear as a pattern in your wins, it is not a signal for your business regardless of how logical it seems.

Mistake 5: Ignoring Closed-Lost Data

Many teams only analyse their wins. The losses are equally valuable because they tell you what to avoid. Without closed-lost analysis, your model will score accounts that look good on paper but will never close. The exclusion criteria from Step 2 are what separate a good model from a great one.

Mistake 6: Not Connecting Signals to Plays

A ranked list of accounts is not an outbound system. If signals fire and nothing happens automatically, you have a research tool, not an infrastructure. Every signal combination should route to a specific play with defined messaging, channel, and timing.

Mistake 7: Treating All Signal Sources as Equal

A custom signal derived from closed-won analysis is not the same as a generic signal from a third-party database. Your scoring model should weight them differently. Custom signals get higher weights because they are more predictive. Generic signals serve as context and supporting indicators.

Mistake 8: Building in Isolation from Sales

The signal model should be built with the sales team, not for them. Sales reps know which deals felt right from the first call and which ones were uphill battles. That qualitative input is essential for identifying patterns that do not show up in CRM fields alone. If your data team builds the model without talking to sales, it will miss the subtle indicators that experienced reps recognise intuitively.

Mistake 9: Expecting Immediate Results

A custom signal model takes 4-12 weeks to build properly. The first version will not be perfect. You will need to calibrate weights, add signals you missed, and remove ones that are not predictive. Plan for a 90-day ramp period where you are testing and refining. The model gets significantly better after two quarterly review cycles.