um, there are signals EVERYWHERE 😥

The Clay Backtest Method for Signal-Based Marketing That Actually Works

Forty-seven signals. Thirteen pieces of tech. Eight hungry sales reps.
And zero clarity on what actually moves the needle.

This isn't uncommon. I've been in this exact conversation with at least a dozen clients in the past six months. If you’re feeling this way too about your own setup right now—you’re far from alone my friend. Everyone's bought into signal-based marketing conceptually, but 90% of implementations are failing because teams are treating it like lead scoring 2.0 instead of discovering what actually works for their specific business.

What I’m seeing is that most companies are copying generic signal frameworks from competitors or "best practice" blog posts (the IRONY in this Substack…I digress). Job changes, funding announcements, website visits, competitor research signals - you know the drill. But what if your buyers don't follow those predictable patterns? What if the signals that actually predict purchase intent for your business are completely different? Odds are high that they are different, actually.

---

👋 Hi, it’s Kaylee Edmondson and welcome to Looped In, my newsletter exploring demand gen and growth frameworks in B2B SaaS. If you’re one of the 119 people that have subscribed since last Sunday, hello! 🥺 Wow, I’m just so honored you’d even consider adding my thoughts to your inbox. I’ll commit to making this everything you expect, and hopefully 10% more sarcasm on the side.

I’ve had a ton of people ask me lately how I got into this wild world of solopreneurship, when I find the time to write, and what things really look like behind the scenes so I’ll start using these sections to be more vulnerable and transparent in the hopes that it helps you in some small way with whatever is next on your journey.

For today, let’s start with the easiest, “when do you find time to write”. And it typically looks exactly like this 👇

It’s 3:30 on a Friday afternoon. I’ve just finished client work for the week. I have 2 hours before I need to be in “mom-mode” for the evening/weekend. And I desperately need to increase my Vitamin D intake for today. No better time while the thoughts are flowing than to jump into writing mode from the porch.

It’s funny how time, and experiences shift your perspective on so many things. You can ask my friend, Chelsea, I used to be known for always saying, “I’m not the words person.” Chronically. I said it every week. I naturally have a very analytical mindset and I’m always rooting decisions and thoughts in logic vs creativity. So naturally to me that meant there was always someone way more qualified in the room to be the “words person”. Don’t get me wrong…this last statement is still true 100% of the time but I found that writing actually helped me further solidify and explore my analytical and logistical thoughts. The better I could write it, the better I could explain it to a boss, a peer, a client, and so on. And so…now I write…weekly on Substack, occasionally on LinkedIn, in the hopes that what I’m processing and trying to solve for might also help you too. Even if just 1%.

Anywho, back to regularly scheduled programming.

---

I’m real big on the practice of backtesting

The truth is I believe your closed won data holds the blueprint for your perfect signal strategy. You just need to know how to extract it.

And that's exactly what I'm going to walk you through how to do today using Clay.

That Execution Gap Is Killing Your Signal Strategy

The strategy part? Most teams have that figured out. They understand the concept of tracking intent signals, they've read the case studies, they've seen the promise of better targeting and higher conversion rates.

But the execution? That's where things are getting a little dicey as the market starts to build this new plane while practically flying 6 other partially built jets let’s be honest.

I see this pattern constantly: Companies implement signal tracking, start collecting data from multiple sources, and then... nothing changes. They're still sending the same generic outbound sequences. Sales is still complaining about lead quality. Marketing is still struggling to prove ROI.

The problem isn't the signals themselves. It's that most companies are flying blind, tracking signals that sound important rather than signals that actually correlate with revenue for their specific business.

Think about it this way: Every company has a different buyer profile, different pain points, different decision-making processes. So why would the same generic signals work for everyone? They wouldn't.

The Clay Backtest Framework

It’s 4 steps. You’ve got this. This framework will help you identify the signals that actually matter for your business, not someone else's business. We're going to work backwards from your wins to identify the patterns that matter most.

Step 1: The Data Foundation Setup

First, we need to get your historical deal data into Clay so we can analyze it.

What Data You Need:

• 12-18 months of closed-won deals (minimum 50 deals for statistical significance)

• 12-18 months of closed-lost deals (ideally 2-3x your won deals)

  • Also before y’all come for me…these are just recommendations to reach stat sig. If you don’t have this much historical data, no biggie, use what you have, just know it’ll be directionally informative.

• Deal information: close date, deal size, sales cycle length, loss reason

• Contact information: name, email, job title, company

• Company information: domain, industry, employee count

• *And the most important bullet point here: And ANYTHING ELSE that you already feel you know is particularly relevant to your business. This could be physical store locations by geo if you’re in the ESG space. Or the number of retailers you sell into if you’re in the CPG space. Or the amount they spend on ads if you’re somewhere in the MarTech space. And so on. You likely have gut feelings or insider info already about what some of these things could be. And the good news is, if you don’t have those data points in your CRM already, odds are, you can enrich them in Clay.

Getting Your Data from Your CRM:

For HubSpot:

1. In Clay, click "Add Data" → "Import from HubSpot"

2. Connect your HubSpot account

3. Select "Deals" as your object type

4. Filter for: Deal Stage = "Closed Won" and Close Date = last 18 months

5. Import the following properties: Deal Name, Amount, Close Date, Associated Companies, Associated Contacts, and any of those magic custom values

6. Repeat this process for "Closed Lost" deals in a separate table

For Salesforce:

1. In Clay, click "Add Data" → "Import from Salesforce"

2. Connect your Salesforce account

3. Create a list in Salesforce first with Opportunity data filtered for closed deals

4. Import this report into Clay with the same property fields

Pro Tip: Start with one table for closed-won and one for closed-lost. We'll analyze them separately first, then compare patterns.

The import process is quick and painless just make sure you have the right permissions. Don't worry if you're missing some data points - we'll enrich everything in the next step.

Step 2: The Signal Mining Process

Now we're going to enrich both datasets with every signal we can possibly gather. This is where Clay really shines - instead of manually researching each account, we'll automate the entire enrichment process.

Enrich what you don’t already have:

Clay's "waterfall enrichment" feature will try multiple data sources to get the most complete picture possible. Here's how to set it up:

1. Company Enrichment Setup:

   • Add a new column called "Company Data"

   • Use Clay's "Enrich Company" action

   • Set up waterfall with: Clearbit → Apollo → People Data Labs → HG Insights (Pro Tip: Structure these from least amount of credits to most amount of credits. It’ll save you some $$$ over time.)

   • This will give you employee count, industry, funding stage, tech stack, etc.

2. Technographic Enrichment:

   • Add column "Tech Stack"

   • Use BuiltWith and HG Insights integrations

   • Focus on: CRM tools, marketing automation, analytics tools, productivity software

3. Intent Signal Enrichment:

   • Add column "Intent Signals"

   • Use Clay’s newly launched Signals to pull in job changes, new hires, job postings, promotions, news & funding, or literally any other custom signal you want to build

4. Engagement Enrichment:

   • If you have first-party data, import website visits, email engagement, content downloads, etc. pull this in via integration as well

Using Clay's AI for Pattern Recognition:

Here's where it gets really powerful. We'll use Claygent (Clay's AI agent) to analyze qualitative data:

1. Add a column called "Company Analysis"

2. Use this Claygent prompt or any other prompt that you’d like honestly (make sure to customize all the inputs to map to your actual fields and knowledge base):

   You are analyzing a prospect account to determine likelihood of purchase based on our enriched data. 
   
   CONTEXT ABOUT OUR BUSINESS:
   - Industry: {{our_industry}}
   - Solution: {{our_solution_description}}  
   - Typical buyer personas: {{our_buyer_personas}}
   - Average deal size: {{our_avg_deal_size}}
   - Average sales cycle: {{our_sales_cycle}}
   
   ACCOUNT DATA TO ANALYZE:
   Company: {{company_name}}
   Domain: {{company_domain}}
   Industry: {{industry}}
   Employee Count: {{employee_count}}
   Funding Stage: {{funding_stage}}
   Recent Funding: {{recent_funding_amount}}
   Tech Stack: {{tech_stack}}
   Recent Hires: {{recent_hires}}
   Job Openings: {{job_openings}}
   Intent Signals: {{intent_signals}}
   Recent News: {{recent_news}}
   Leadership Changes: {{leadership_changes}}
   Growth Indicators: {{growth_indicators}}
   Competitive Tools: {{competitor_tools_used}}
   
   ANALYSIS REQUIRED:
   Based on ALL the data above, provide a structured analysis:
   
   1. WIN PROBABILITY FACTORS (Score 1-10 for each):
      - Strategic fit with our solution
      - Financial capacity and urgency
      - Technology readiness and adoption patterns  
      - Organizational change indicators
      - Competitive displacement opportunity
      - Timing alignment with business priorities
   
   2. SPECIFIC EVIDENCE:
      For each factor above, cite specific data points that support your scoring
   
   3. RED FLAGS:
      Identify any signals that suggest this account is unlikely to buy or would be a poor fit
   
   4. ENGAGEMENT STRATEGY:
      Based on your analysis, recommend:
      - Primary contact persona to target
      - Key pain points to emphasize
      - Optimal timing for outreach
      - Competitive differentiation angles
   
   5. OVERALL LIKELIHOOD SCORE: 1-100
   
   Format your response as structured data that can be easily analyzed for patterns.

3. Run this across both your won and lost deal datasets

Fingers off the keyboard. Let Clay do it’s magic.

Step 3: Pattern Recognition & Signal Validation

Now for the detective work. We're going to identify which signals appear significantly more often in your closed-won deals than your closed-lost deals.

Creating Your Comparison Analysis:

1. Merge Tables Natively:

   • Go to your closed-won table

   • Actions → "Append rows from table" → Select closed-lost table

   • Clay automatically combines them into one master analysis table

2. AI Pattern Analysis Setup:

   • Add column "Pattern_Analysis" with your Claygent prompt

   • Add column "Signal_Correlation_Scores" to extract structured insights

Some Wild AI Signal Analysis

Replace simple correlation formulas with pattern recognition:

Column 1: Comprehensive Pattern Analysis

You are analyzing a {{Deal_Status}} deal to identify predictive patterns.

DEAL OUTCOME: {{Deal_Status}} ({{Deal_Size}} deal, {{Sales_Cycle_Length}} days)
LOSS REASON: {{Loss_Reason}} (if applicable)

COMPANY DATA:
- Company: {{company_name}} ({{employee_count}} employees)
- Industry: {{industry}} | Funding: {{funding_stage}}
- Tech Stack: {{tech_stack}}
- Recent Hires: {{recent_hires}}
- Job Openings: {{job_openings}}
- Growth Signals: {{growth_indicators}}
- Intent Data: {{intent_signals}}

ANALYSIS REQUIRED:
1. PREDICTIVE SIGNAL STRENGTH (Rate 1-10 for each):
   - Company size/growth trajectory fit
   - Technology stack alignment
   - Hiring patterns indicating need
   - Funding/budget capacity signals
   - Intent/engagement strength
   - Competitive situation

2. PATTERN COMBINATIONS:
   What combination of 2-3 signals best predicted this {{Deal_Status}} outcome?

3. TIMING FACTORS:
   Which signals indicated optimal or poor timing?

4. HIDDEN CORRELATIONS:
   What non-obvious data points might have predicted this outcome?

Respond in structured JSON format for analysis.

Column 2: Signal Score Extraction Use Clay formulas to extract structured scores from the AI analysis:

REGEX_EXTRACT({{Pattern_Analysis}}, "Technology stack alignment: (\d+)")

Advanced Pattern Recognition Analysis

What to Look For (Beyond Simple Percentages):

Instead of just "75% of won deals had Signal X," you'll discover patterns like:

Multi-Variable Patterns:

• "Series B companies + recent VP Sales hire + Slack usage + 20%+ growth = 89% win rate"

• "Technical evaluation signals + budget authority contact + competitor research = 72% win rate"

Timing Correlation Patterns:

• "Job posting → website visit within 14 days = 3x higher close rate"

• "Funding announcement → demo request within 30 days = 45% faster sales cycle"

Sequential Signal Patterns:

• "Leadership hire → tech stack research → pricing page visit = buying journey progression"

• "Competition research → case study downloads → team expansion = readiness indicators"

Negative Signal Patterns:

• "Budget research without stakeholder expansion = 78% loss rate"

• "Technical interest + procurement involvement early = 65% longer sales cycle"Setting Up Signal Correlation Tracking:

Real Example from a Recent Client:

A MarTech company discovered these validated signals:

• Signal #1: Companies using both Salesforce AND a customer success platform (78% of wins vs. 23% of losses)

• Signal #2: Recent hiring of VP-level revenue operations role (82% of wins vs. 31% of losses)

• Signal #3: Engagement with ROI-focused content in past 90 days (71% of wins vs. 19% of losses)

These became their "golden signals" - any ICP fit prospect showing 2+ of these signals got tier 1 treatment.

Step 4: Implementation & Automation

Now we build signal detection. This is where Clay becomes your competitive advantage - automatically identifying high-intent prospects based on your validated signal patterns.

Building Signal Scoring:

1. Create Your ICP Scoring Table: Set up a new Clay table for prospect scoring

2. Import Your Target Account List: This could be from your CRM, a purchased list, or Clay's company finder

3. Set Up Signal Detection: Create columns for each of your validated signals

4. Build Your Scoring Formula: Weight signals based on their correlation strength. Clay has out of the box scoring functionality built in. When you select Add Column —> Add Enrichments —> Score Row in Clay

You can definitely make your weighting straightforward or complex. Start simple and build in more complexity over time. Here's a simple scoring formula:

{{Signal_1_Score}} * 3 + {{Signal_2_Score}} * 2.5 + {{Signal_3_Score}} * 2

Automating Signal Detection:

Set up Clay automations to:

• Check for new prospects weekly, daily, whatever works best

• Enrich them with your validated signals

• Score them based on your formula

• Trigger different campaigns based on score ranges

CRM Integration for Sales:

Push your signal scores back to your CRM so sales + marketing can prioritize:

1. Set up Clay → Salesforce/HubSpot integration

2. Create custom fields for signal scores

3. Set up automated lead routing based on scores, geos, etc.

4. Create alerts for high-scoring prospects

The Technical Stuff

Let's get into the specific Clay setup so you can replicate this.

Clay Plan Requirements:

• Minimum: Clay Explorer ($349/month) for CRM integrations and advanced enrichments

• Recommended: Clay Pro ($800/month) for unlimited automations and higher credit allowance

• Credit Budget: Plan for about 5-10 credits per prospect analyzed (varies based on enrichment depth)

Common Clay Setup Mistakes to Avoid:

1. Credit Burn: Don't run enrichments on every possible data point - focus on the signals that matter. And make yourself a hierarchy structure so that any account/contact you’ve enriched in one table, you don’t accidentally re-enrich for the same information again in another table. I know this mistake far too well sadly.

2. Data Quality: Clean your CRM data before importing - bad data in, bad analysis out

3. Prompting, then prompting again: Don’t just blindly run your first prompt for your entire table. Prompt, rigorously review that preview, maybe even build the confidence to test it on the first 10 rows. And if it sucks, which the first few always do, prompt again. I’ve found using voice dictation has really helped wonders with my prompting game. Try a tool like Willow 💡

4. Sample Size: Try to avoid drawing concrete conclusions from less than 50 closed deals (signal vs noise)

5. Overfitting: Don't create a model so specific it only works for past deals

This Creates a Competitive Moat

This approach gives you a sustainable advantage because:

Your Signal DNA is Unique: Every company's signal patterns are different. Your competitors can't easily copy what you discover because it's based on your specific customer data.

Continuous Improvement: As you close more deals, you can refine your signals. Your competitive advantage actually gets stronger over time.

Operational Efficiency: While competitors are still manually researching prospects or using generic signals, you're automatically identifying and prioritizing the highest-intent accounts.

Sales and Marketing Alignment: Both teams are working from the same validated data about what actually predicts success.

Or The Contrarian Take: Why You Might Want to Ignore This Advice Altogether 🤷🏻‍♀️

Let me argue against my own framework for a minute, because there are legitimate reasons why this approach might not be right for your company.

The "Good Enough" Argument: Maybe your current approach is working fine. If you're hitting your numbers with existing methods, the time investment required for this level of analysis might not be worth it. Sometimes "good enough" really is good enough.

The Simplicity Argument: There's something to be said for keeping things simple. Tracking 3-5 basic signals that everyone understands might be more effective than a sophisticated system that's too complex for your team to manage properly.

The Speed Argument: This analysis takes time. If you're in a fast-moving startup where priorities change monthly, you might be better off with more agile, less data-dependent approaches.

The Resource Argument: Clay isn't cheap, and this approach requires dedicated analytical resources. Smaller teams might get better ROI from investing those resources directly in more outbound volume or sales hiring.

The Innovation Argument: Relying too heavily on historical data might blind you to new opportunities. If your market or product is evolving rapidly, what worked for past deals might not predict future success.

But here's why I still believe this approach wins for most B2B companies:

The competitive landscape is getting more crowded. Generic outbound strategies are getting less effective. Every medium is becoming more crowded than ever. The companies that win are going to be the ones with unique data advantages (and insanely talented and amazing humans, and I can’t stress that enough…but a topic for a different day).

---

Here's to growth and building some cool stuff in Clay!
See ya next week,
Kaylee ✌

Want more frameworks like this? Reply and let me know what GTM challenge you're facing - I read every response and often turn them into future newsletter topics (anonymized of course).