GTM Engineering Best Practices

Best practices in GTM Engineering with the Head of GTM Engineering at the company that invented the domain: Clay’s Osman Sheikhnureldin

GTM engineering is NOT cold outbound done better.

It’s about using technology to eliminate the constraints that stop you from growing even faster.

Most founders get it wrong - GTM engineering is about way more than sending more emails faster.

In the latest episode of the focal podcast, I sat down with Osman Sheikhnureldin, the Head of GTM Engineering at Clay (the company that invented it) to walk through how to actually get the most out of GTM Engineering.

Plus I also brought on Nilo Rahmani (co-founder / CEO of Thoras AI) and Panos Papageorgiou (Co-founder of Keragon) to work through real GTM engineering examples live on the pod.

Here are 6 lessons from my conversation with Osman, Nilo, and Panos - listen to the full episode here:

1. The list matters more than the message.

This is counterintuitive.

Founders obsess over their email copy, A/B testing subject lines and tweaking their calls to action. However, if you’re reaching out to the wrong people or the right people at the wrong time, it doesn’t matter how good your message is.

Here’s why: Cold outreach is interruptive by nature.

You’re reaching someone who wasn’t thinking about you or your product. The only way that works is if they happen to have the exact problem you solve, right now. Not six months ago. Not next year. Now.

This seems obvious, but most founders get it backwards - they pour energy into perfecting their pitch while sending it to a generic list scraped from LinkedIn.

At scale, cold outreach works with tiny response rates: 1-2% if you’re good. Big companies can make that math work because they have the resources to process thousands of lukewarm leads.

But as a small startup, you can’t afford spray and pray like that. You don’t have the team to follow up on hundreds of “maybe interested” responses.

Instead, you need a much smaller list of people who are almost certainly in-market right now - and then your response rate can be dramatically higher.

The list isn’t just “find the right job title.” It’s about identifying who has your problem acutely, at this moment, based on what’s actually happening in their business.

That requires a completely different way of thinking about targeting - it’s highly effective if you get it right.

2. Static ICPs are dead.

To double down on what I mentioned above: The way most founders think about their Ideal Customer Profile is too static. “Marketing managers at early-stage startups who use HubSpot.” That tells you almost nothing about whether they need what you’re selling right now.

Think about what that ICP actually gives you: a job title, a company stage, and a tool they use.

But it doesn’t tell you anything about what that person is trying to accomplish this quarter. It doesn’t tell you if they just got budget approved for a new initiative, or if they’re drowning in a problem you can solve, or if they’re actively evaluating solutions.

The shift you need to make is from static attributes (ICP) to jobs-to-be-done of your target customers:

• What is this business actively trying to accomplish?

• What initiative just kicked off?

• What pain became acute last week?

Consider the difference:

• Static ICP: “VP of Sales at Series A startups.”

• Jobs-to-be-done: “A company that just hired a new VP of Sales, has job postings up for five SDRs, and recently adopted Salesforce.”

The second version tells you a story. This company is building out their sales motion right now. They’re in the middle of it. That’s when they need help and when they are most receptive to your outreach if what you’re selling them can solve their current problems.

The data points you’re looking for aren’t demographics. They’re evidence of initiatives, transitions, and pain points that are live right now.

3. Single signals lie. Combined signals tell stories.

One signal rarely tells you much. A company hired a VP of Sales - so what? Thousands of companies did that this month.

But when you combine signals, they paint a picture.

That’s a big part of what GTM Engineering is about: It isn’t just to collect data points about a prospect. It’s to construct a story about what that business is actually doing right now.

The example that stuck with me:

• A founder who is selling to European food companies expanding to the US.

• How do you find companies in that exact situation? They looked for companies with 100+ employees but only one US-based employee hired in the last six months. Which suggests that this is a company that’s established in Europe but just starting their US presence.

• Add to that: recent ad spending in the US market. And a new US-specific page on their website.

None of those signals alone means much. Together, they tell you exactly where this company is in their expansion journey - and exactly when they need help.

Combining these signals into a story then not only allows you to target them better but you can also send a way more personalized email as you understand where they are in their journey - thereby, you’re demonstrating that you understand their situation better than they expected anyone to.

In summary, a big part of GTM Engineering is being creative around how to best find and combine signals / data points in a way that allow you to construct a story.

You’re looking for combinations of publicly available data - job postings, website changes, ad spend, technology adoption, press releases, social posts, etc - that together reveal what a company is prioritizing right now.

4. Ask the magic wand question.

As you’re getting started with GTM Engineering, as yourself (before you think about data sources or tools):

If you had a magic wand and could appear in front of any prospect at the exact moment something became true about their business, when would that be?

Don’t constrain yourself by what data you think is available. Start with the ideal scenario, then figure out how to approximate it.

For Keragon who was on the podcast, the answer was:

The 30-90 day window after a hospital rolls out a new electronic health records system. That’s when their old workflows break and the new system is live, but it doesn’t talk to their other tools (yet). Meaning data that used to flow automatically now requires manual copy-paste. Staff are drowning. The pain is acute. A perfect scenario for a company like Keragon that is building the automation platform for Healthcare (instantly connect all your healthcare software to build HIPAA-compliant automations, without code).

Their answer to the magic wand question is incredibly specific. Which is why it allows for great targeting:

Once you know that’s your magic moment, you can get creative about finding signals that indicate it’s happening. Maybe it shows up in press releases. Maybe hospitals announce EHR transitions to patients. Maybe you can detect new subdomains or API endpoints that indicate which system they’re running.

The magic wand question forces you to think about timing and context, not just company attributes.

5. Founders have a special advantage - if they use it.

Here’s something the GTM engineering crowd doesn’t talk about enough: founders who come from the domain they’re selling to have a massive edge.

When you’ve lived the pain, you don’t describe it abstractly. You describe it viscerally.

You know what it feels like when the pipeline breaks at 2am. You know the specific dread of seeing red lights in Sentry while you’re on vacation. You know the awkward conversation with your CEO when the dashboard is wrong.

That specificity is what makes prospects think “this person actually gets it.” A sales rep who’s done market research will say “infrastructure reliability is important.” A founder who’s lived it will say “you’re probably getting paged at 2am and spending the first hour just figuring out which service is actually broken.”

No amount of market research substitutes for having been in the trenches yourself. If you have that advantage, use it in your messaging. It’s your unfair edge.

6. Commoditization makes human taste more important, not less.

Here’s the paradox: as AI makes it easier to send personalized emails at scale, human creativity becomes more valuable, not less.

When everyone can send a thousand cold emails per day, that tactic stops working. The emails start to blend together. Inboxes are flooded with messages that are technically personalized - they mention your company name, maybe your recent funding round - but feel generic because they all follow the same playbook.

What breaks through is genuine creativity and taste in how you approach problems. The founders who think they can plug in some generic AI SDR tool and watch pipeline magically appear will be disappointed. That promise hasn’t borne out, especially for smaller companies. The tools are table stakes now. What matters is what you do with them.

The question isn’t “how do I automate outreach?” It’s “what creative insight about my prospect can I surface that no one else is seeing?”

That’s why the story mentioned above is so powerful - if you get it right, you can instantly connect with the recipient / they feel spoken to.

A low-effort, generic message is an annoyance. We all get them. “Hi, I noticed you’re in the tech space and thought you might be interested in...” Delete. But a hyper-targeted message that demonstrates deep understanding of their specific situation feels different. It feels like someone actually did their homework. It feels like it might be worth reading.

7. It takes months, not weeks.

Founders want GTM engineering to produce results in a month. It won’t.

This isn’t something you can wing. You need to deeply understand your customer’s pain - not at the persona level, but at the “what are they struggling with this week” level.

You need to translate that into data signals you can actually detect. You need to figure out how to source that data, often from multiple places. You need to craft messaging that resonates with people in that specific situation. Then you need to test, see what works, and iterate.

Each step takes time. And the steps compound - you can’t write good messaging until you understand the signals, and you can’t find the signals until you understand the pain.

If someone tells you they can set up your GTM engineering in a week, they’re selling you a template. Templates are what everyone else is using. You need something better.

8. You’re never done.

The market changes. Your product evolves. What worked six months ago stops working. The signals that once predicted high-value customers shift.

GTM engineering is a constant process of iteration. You hear something from sales about a pattern they’ve noticed - maybe that deals close faster when the prospect has a certain kind of organizational complexity - and you try to turn that into a discrete data point. You discover that companies who’ve recently done acquisitions tend to have messy tech stacks, so you figure out how to detect that.

For example, at Clay, they built a simple machine learning model to predict contract value based on dozens of features about each account.

One of the useful outputs: they can see which features most strongly predict value. That tells them where to focus. But it’s never finished. They’re constantly adding new signals based on what they learn from sales conversations.

Meaning GTM engineering isn’t a project you complete. It’s a discipline you practice. You can never stop learning.

---

Enjoyed reading this?

Then check out my conversation on the focal podcast with Osman Sheikhnureldin, Head GTM Engineering at Clay (the company that invented GTM Engineering)

Youtube | Apple Podcast | Spotify

---

Recently started a company or thinking about it?

At focal, we’re technical, AI native builders’ first choice for their first check.

We lead their first round at the very start with up to $1M. Often before they even write their first line of code.

Reach out.