Founder field report
I Sent 200 Cold Emails. Two Companies Wanted to Try My Product. I Still Screwed It Up.
I thought the hard part of cold outreach was finding leads and writing convincing emails. Turns out, I was optimizing the wrong thing.
In April and May 2025, I ran a cold email campaign for my proxy infrastructure startup.
I sent roughly 200 emails.
Most went nowhere. Plenty bounced. Some reached companies that probably had no reason to buy what I was selling.
But two companies responded with genuine interest.
One offered to arrange a call. Another explicitly asked for a test account.
And I still managed to mess up one of the best opportunities in the entire campaign.
Looking back, that failure taught me more about sales than any improvement to my email copy could have.
The campaign looked productive. It wasn't particularly effective.
I was doing what a lot of technical founders do when they start selling.
I used tools to find companies, research prospects, generate personalized messages, and send outreach.
The workflow could produce plenty of activity. But activity and progress turned out to be very different things.
My rough count suggested that around 20% of the emails had terminal delivery failures. That alone was a problem.
The bigger issue was targeting.
I started with a reasonable idea: find companies that depend on web scraping, data collection, or proxy infrastructure.
Those companies might have a real reason to care about what I was building.
But as the campaign continued, the definition of a potential customer became increasingly loose.
A company used APIs? Maybe it needed proxies.
A company worked with data? Maybe it did scraping.
A company had a technical product? Surely I could find some angle.
And with AI helping write the messages, finding an angle became dangerously easy.
AI can personalize an email without understanding why someone should buy
This was probably the most interesting failure.
AI is quite good at turning company research into something that sounds personal.
It can mention a product, identify a technical detail, and produce a polished paragraph explaining why your offering might matter.
But there's a gap between knowing something about a company and knowing that it has a problem worth solving.
A company might use APIs without needing residential proxies.
A company might collect data without having a proxy-related bottleneck.
A company might operate at enormous scale without being interested in replacing its existing infrastructure.
The generated emails sometimes crossed that gap without evidence.
They sounded informed, but the underlying buying premise was speculative.
Some personalization was technically specific and still irrelevant.
I was effectively using AI to make weak leads sound stronger.
And because the resulting messages were fluent, that problem wasn't always obvious.
A convincing reason to send an email is not necessarily a good reason to contact the company.
Then two genuinely interesting companies replied
Among all those emails, two responses stood out.
A search-infrastructure company was interested in the product and offered to set up a call. We later spoke, and its CEO followed up to connect directly.
That made sense. Search infrastructure is a much more natural fit for proxy technology than a random software company with an API.
A web-scraping and data-collection company was another strong match.
Its business depended on collecting data from websites—the kind of work where proxy infrastructure can matter.
My outreach was forwarded internally, and the company explicitly asked for a test account.
Not a vague compliment. Not an automated response.
A request to actually try the product.
These were precisely the kinds of companies I should have concentrated on in the first place.
Out of roughly 200 emails, I had found two meaningful opportunities.
What happened next is the embarrassing part.
Someone asked to try the product. I sent them to the waitlist.
The second company asked for a test account.
Instead of immediately helping them evaluate the product, I directed them to the public waitlist.
Think about that for a second.
I was a founder sending cold emails because I needed people to try my product.
A relevant company responded and asked to do exactly that.
And I gave them another step to complete.
I should have treated that reply as a high-priority opportunity: arrange access, understand what they wanted to test, and make the evaluation as easy as possible.
Instead, I responded as though a genuinely interested prospect were just another signup.
I can't claim that handling it differently would have resulted in a customer. Maybe the product wouldn't have met their requirements.
But that wasn't even the question yet.
The immediate opportunity was to get a serious potential customer testing the product.
I had already done the difficult part of getting their attention, and then added friction at precisely the wrong moment.
The real problem wasn't email generation
For a while, I thought better outbound meant better automation.
Find more leads. Research them faster. Write more personalized emails. Improve deliverability. Send more.
Those things matter.
But they don't solve the most important problem.
At every stage of outreach, somebody—or something—has to make a decision.
And a bad decision at any stage can make everything afterward pointless.
Looking back, there were four questions I should have asked consistently:
- Is this actually a good potential customer? Not somebody who vaguely operates in the same industry, but a company with a credible reason to need the product.
- What do I actually know about them? Which facts are backed by sources, and which parts are assumptions dressed up as personalization?
- Why contact them now? Is there evidence of a relevant problem, change, or opportunity—or am I inventing urgency because I want to send an email?
- What's the right next action? Should I reach out, research further, hold off, arrange a call, or give somebody who explicitly requested it a test account?
That fourth question was where I made my most painful mistake.
It's also the question that many outbound workflows barely address.
They're good at moving contacts through a sequence.
They're less good at recognizing when the sequence should stop and a different decision needs to be made.
Why this experience changed what I wanted to build
This experience became part of the thinking behind the lead-review product I'm building now.
There are already powerful tools for finding contacts, enriching records, researching companies, and automating outreach.
Existing prospecting and enrichment tools can help teams assemble and work through large prospect lists.
I don't think the world needs another tool whose primary purpose is to send more emails.
What interests me is the decision layer between finding a prospect and taking action.
Before contacting somebody, can we establish that the company actually fits?
Can we distinguish evidence from speculation?
Can we identify unsupported claims in a message?
Can we catch duplicate records, contact restrictions, and other reasons an otherwise plausible prospect shouldn't proceed?
And can we make the remaining uncertainty visible instead of hiding it behind a confident AI-generated recommendation?
That's the direction I'm pursuing.
The goal isn't to have AI confidently approve every lead. Sometimes the most useful result is a clear explanation of why a lead isn't ready.
It's about making the reasoning and evidence reviewable before an agent—or a human—acts.
The lesson I wish I'd learned before sending those emails
If I ran that campaign again, I wouldn't begin by asking how to send 200 emails more efficiently.
I'd start by identifying the ten or twenty companies with the strongest reasons to care about the proxy product.
I'd spend more time checking those reasons.
I'd be much more skeptical of AI-generated personalization.
And when somebody asked for a test account, I'd make sure the next step was actually getting them one.
None of this guarantees more sales.
But it avoids some remarkably unnecessary mistakes.
The most frustrating part of that campaign wasn't the unanswered emails.
It was discovering that I'd managed to find genuine interest and still failed to handle it properly.
I didn't need another tool to send more emails. I needed a system that could help me make fewer bad decisions.
That's what I'm trying to build now.