do you trust the list?

Hey Predictable Revenue community,

Rev Roundtable Update: Our founder revenue roundtable is happening again next week (April 11th at 9:30am PT) and we have a few more spots left, hit me up if you’d like to join us. This is specifically for founders under $1m in revenue that are looking for help unblocking a particular revenue obstacle. We pick 3-4 companies/struggles as a group and share what’s worked for us to help them find a way forward. Hit reply if you’re interested.

I built a list this week that gave me that cold, sinking feeling – the realization I’ve been approaching list building with a level of risk recently that borders on negligent. Like ‘driving your GTM strategy without a seatbelt’ dangerous.

Here’s the gist:

I was working with a founder trying to answer a critical question: Is there enough runway in their current market segment to hit their revenue goals, or do they need to expand into new verticals? Standard top-down research reports showed their industry had roughly twenty thousand companies and Apollo only showed ~600 companies that specifically met the employee range we were looking for.

This founder wants to move slightly upmarket, shifting focus from smaller companies to those with a minimum 50 employees and max 200. Why? Because these larger companies derive significantly more value from their software, meaning bigger deals, better adoption, and longer client lifetimes.

The Initial Plan & The Nagging Doubt

We started by building a bottoms-up sales development forecast model. (Hit reply if you'd like my template for this – happy to share). This model helped us map out how long it would take to penetrate the assumed market size and what profitability could look like based on various inputs (conversion rates, sales cycle length, etc.). It felt like a solid start.

But something was nagging at me.

The entire model – the hiring plan, the GTM investments, the sales commission structure – hinged precariously on one number: the total number of accounts with 50-200 employees. My client was about to bet a significant chunk of their next year's strategy and budget on this figure. Any major error would ripple through the entire company.

We had cross-referenced our target number with data from a couple of reputable providers, and their figures generally aligned. Yet, my spidey sense was tingling. I couldn't shake the feeling that trusting these numbers implicitly for such a high-stakes decision wasn't good enough.

Digging Deeper: Validating Assumptions the Hard Way

We agreed that a decision this fundamental deserved extra diligence. It was time to validate our market size assumptions from the ground up. Here’s what I did:

1. Data Collection via Clay: I built a Clay worksheet leveraging the Google Maps API. The goal was to identify any company whose online presence included keywords relevant to the founder's specific niche within Industry N.

2. AI-Powered Filtering: Crucially, I then created an AI classifier prompt within Clay. This prompt instructed an LLM to analyze each company's description and determine if it was truly the right type of company for my client – not just tangentially related.

(Sidebar: Why Standard Data Providers Can Fall Short Here)

This process highlights a common challenge with third-party data providers. They often get criticized for data quality, but it's partly by design. Their incentive is usually to maximize the volume of data matching broad criteria – they don't want to risk leaving out a potential lead for their customers. In casting this wide net, they inevitably pull in companies that technically fit the filters (industry, employee size) but aren't actually a good fit for your specific solution or value proposition. The issue isn't the raw data itself, but the lack of sufficiently granular filters to precisely define your ideal customer profile based on qualitative factors.

The Alarming Results

The outcome of our deep dive was both validating and alarming.

• The Good: The total potential number of companies in the space loosely matched the top-down reports. The market was roughly that size on paper.

• The Bad: The number of companies that were a good fit and within the target employee range was nearly 50% lower than the initial data suggested.

The "Near Miss" and Its Potential Fallout

I couldn't shake that "near miss" feeling – like watching a potential car crash unfold and realizing a slightly different decision could have put you right in the middle of it.

Fortunately, my intuition kicked in, and the client readily invested the extra time to get the foundation right. But let's imagine we hadn't:

1. Over-Hiring: They would have built their hiring plan based on a market twice as large as reality, likely recruiting double the necessary salespeople.

2. Missed Quotas & Morale: Sales reps would have struggled immensely, facing a scarcity of viable accounts. At best, the team might have hit 50% of their target, leading to frustration, demotivation, and eventual turnover.

3. Wasted Resources: Think of the management overhead spent hiring people they didn't need, the months spent managing underperforming (through no fault of their own) reps, and the painful process of rightsizing the team later.

4. Strategic Risk: This wouldn't just burn cash and time; it could have jeopardized their ability to hit milestones crucial for their next funding round or achieving profitability. Burning through resources based on flawed data can quickly deplete a startup's limited "matches." While perhaps not fatal in this specific instance, it would have been dangerously close – all stemming from one unchecked assumption in a spreadsheet.

The Real Hero: Accessible AI + Clay

The real star here is 03-mini. Before some of the more recent models, having an AI look at all the relevant details and make an assessment that I would trust my GTM program felt like science fiction. We used to have a big team of folks to manually process lists like these, it was so expensive that it only made sense to run for really small lists of really high value targets. With 03 mini + Clay, it’s so reliable and cheap that it’s almost irresponsible not to do this for every list I’m running. 

Before LLMs, having an AI reliably analyze qualitative details (like a company description) and make a nuanced judgment call you could confidently base your GTM strategy on felt like science fiction, or at least were prohibitively expensive.

I've built top-down/bottoms-up forecasts before, often relying on subscriptions to expensive data vendors. While those provide a starting point, anyone who's run serious outbound campaigns knows list quality degrades quickly. It's like a Google search – page one looks promising, but by page three or four, the relevance often nosedives.

Applying This Yourself: AI-Powered List Qualification

The only robust way to combat this decay and ensure high list quality is to have an AI individually assess each potential account before it makes it into your strategic plans or your reps' outreach sequences. The best way to do this is to use Clay to inject the Company Name and Description variables into a prompt and run it at scale. It’s the same idea as sending a cold email to “Hi First Name” but instead of sending an email, you’re injecting variables into the prompt.

A simple method involves:

1. Pulling company names and descriptions (e.g., from Google Maps, LinkedIn, or other sources) in Clay.

2. Feeding the company name & description into an LLM via a "classifier" prompt.

3. Filtering based on the output of the classifier

4. Sourcing contacts for “good fit” companies

This is excellent for basic categorizations like "Software vs. Services," "B2B vs. B2C," or, as in my example, "Is this company actually in my specific niche based on what they do?"

Here’s a simple Clay template that does just that.

To use it:

1. Sign up for Clay (if you haven't already).

2. Add this template to your account.

3. Add your OpenAI API key in Clay's integrations settings.

4. Adjust the classifier prompt for your business.This is key. You need to tell the AI exactly how to judge the companies based on the data it will receive (in this basic template, primarily the company description).

How to Create an Effective Classifier Prompt (Using ChatGPT or similar):

You can ask an AI like ChatGPT to help you draft the prompt. Here’s a prompt I used yesterday:

“write a classifier prompt for me, the inputs will be company name and linkedin description. I'm looking for companies that <blank_a> but not <blank_b>. Here are 3 examples of companies that <blank_a> and 3 examples of companies that <blank_b>. 

Key Tips for Your Prompt:

• Be Specific: Clearly define what makes a company a good or bad fit based on information likely found in a description.

• Use Examples: Provide clear examples of both good and bad fit descriptions to guide the AI.

• Constrain the Output: Tell the AI exactly how to format its answer (e.g., just 'Good Fit' or 'Bad Fit'). This makes it easy to use the output in tools like Clay.

The Takeaway

Don't let seemingly solid top-down numbers or basic list filters lull you into a false sense of security when making critical GTM planning decisions. Validate your actual addressable market with a level of granularity that reflects your ideal customer fit. Thanks to accessible AI, this deeper validation is no longer a luxury; it's a crucial step to de-risk your strategy and build your growth plans on solid ground.

Have you faced similar challenges with market sizing or list building? Hit reply – I’d love to hear your experiences.

Collin

PS - if you made it this far, let me know if you want more geeky-gtm posts like this or less. Just hit reply with “more” or “less”.

PPS - if you’re struggling with the sheet, I can help but my time costs $ 😉. 

PPPS - I’m thinking of switching from publishing on Fridays to Wednesdays because newsletters get better engagement during the week. Good idea? Bad? I’d love to hear from you.