The math of "we tried that"
A founder told me last month he'd tried three channels in six months and none of them worked. LinkedIn outreach. Cold calling. Cold email. All of them "tested." All of them "not working." He wanted to know what to try next.
I asked him one question: how many sends had he done on his best-performing channel in the last 90 days?
He thought about it. Then he said maybe 80.
Eighty sends. Split across three or four different messages, sent at various times over about two months, with the ICP shifting each round because he was learning as he went.
That's not three failed channels. That's zero tested channels. He hadn't produced enough signal on any one of them to know whether it worked or not. He was declaring "cold email is dead" based on statistical evidence a first-year statistics student would laugh at.
This is one of the most common patterns I see in early-stage B2B SaaS outbound, and it produces a very specific failure mode. Founders spend six to twelve months rotating through channels, generate almost no signal on any of them, and end up in month twelve genuinely believing that outbound doesn't work for their business. The problem isn't that outbound doesn't work. The problem is that they've never actually run outbound long enough to find out.
What "trying" a channel actually looks like
Here's what most founders mean when they say they've "tried" a channel.
They picked a channel because they read something on LinkedIn about it. They set up the tooling over a few days. They wrote a message. They sent it to 30 or 40 people over a couple of weeks. They got two replies, or one, or zero. They looked at their dashboard on a Friday and decided the channel wasn't working. They moved on to the next channel and repeated the cycle.
This is not testing a channel. This is auditioning a channel. And auditions produce theater, not data.
Real testing looks different. Real testing means picking one channel, one ICP, one message variant, and running it consistently until you've produced enough sends to distinguish signal from noise. In cold outreach, that number is at minimum 500 sends per controlled variant. Some categories need 1,000.
Fewer than 500, and your reply rate could be 2% or 12% and you wouldn't be able to tell from the data. Fewer than 500, and one enthusiastic reply from a founder-friend who happens to be in your ICP will move your metrics enough to feel like a signal when it isn't one. Fewer than 500, you don't know anything. You have a vibe.
Vibe-based outbound decisions are how founders end up in month twelve concluding that outbound doesn't work for their space.
The variables that ruin your comparisons
The other half of the problem: even when founders send enough messages to have real data, they've usually changed too many things during the run to know what worked.
I've seen founders do this exact sequence in a single month:
- Week 1: new subject line, testing it against the old opener
- Week 2: new opener, testing it against the new subject line
- Week 3: different ICP because week 2 got weird results, plus a shorter question
- Week 4: different send time and a rewritten first line based on a comment someone made on LinkedIn
By the end of the month, they've sent 400 emails. They've had 12 replies. And they can't tell you which of the five changes moved the needle, because they changed five things simultaneously.
Their conclusion: "we got some replies but nothing consistent, moving on to LinkedIn."
The right way to test is to hold everything constant except one variable at a time. Subject line changes: keep everything else the same for 200 sends. Opener changes: keep everything else the same for 200 sends. This feels slow. It is slow. It also produces actual answers instead of stories about answers.
Founders who don't do this end up having the same conversation six months apart. "We tried cold email but the reply rates were inconsistent, so we're trying LinkedIn." Six months later: "We tried LinkedIn but the reply rates were inconsistent, so we're trying paid ads." The reply rates weren't inconsistent because the channel didn't work. They were inconsistent because the founder was changing five things at once.
System, not experiment
The reframe that gets founders out of this trap is a single word swap. Stop thinking about outbound as an experiment. Start thinking about it as a system.
An experiment is short. It has a hypothesis, a run period, a conclusion, and then it ends. If it worked, you're happy. If it didn't, you move on to the next experiment. Most founders run outbound as an experiment: "let's try LinkedIn for a month and see if it works."
A system is different. A system is something you're operating continuously. You have a repeatable process. You have a control message. You have measured baselines. You have a next variant queued up to test against the control. And you keep running it, week after week, whether or not this week's numbers were exciting.
A system produces compound learning. Month one you learn what your baseline reply rate is. Month two you learn which of two subject lines wins. Month three you learn which of two openers wins on top of that subject line. By month six, you have a message that's been through five controlled iterations and outperforms the average founder's best guess by a significant margin.
An experiment produces months of activity and no compound learning. You keep running fresh experiments because you keep discarding what you built.
The founders who make outbound work operate systems. The founders who don't, run experiments. The two groups often work equally hard. Only one of them ends up with a channel that works.
Why technical founders especially fall into this
I see this pattern most sharply in technical founders. There's a specific reason.
Technical founders are trained on debugging cycles that are short. You write code, you run it, you see the error, you fix it, you run it again. The feedback loop is seconds or minutes. If a fix doesn't work, you know within minutes and try something else.
Outbound feedback loops are days or weeks. You send 100 emails Monday, you get maybe 3 replies by Thursday, and you can't statistically distinguish those 3 replies from noise until you've run three more weeks of similar volume. The debugging instinct (if it's not working after a few tries, change something) is exactly wrong at this timescale. Every change resets the clock. Every change confounds the previous data. Every change turns weeks of accumulated learning into weeks of new noise.
The technical founders I've watched succeed at outbound are the ones who consciously override this instinct. They set a rule for themselves: I am not changing anything about this sequence for 30 days. Then they run it. Then they change one variable and run it another 30 days.
The ones who fail are the ones who "keep iterating" the way they iterate on code, and end up with six months of activity that didn't compound because they never held anything constant long enough for the results to stabilize.
When to actually pivot channels
There is a right time to pivot channels, and it's later than you think.
The signal that a channel isn't going to work is not "I sent 100 emails and got 2 replies." That's noise.
The signal is: you've sent 500 to 1,000 sends on a controlled message, you know your reply rate is stable at under 3%, you've tested three or four meaningfully different message variants against the same ICP, and none of them cleared 5%. Plus you have reason to believe the ICP itself is correct.
If all of those are true, the channel is probably wrong for your business or your buyer isn't reachable that way. Pivot with confidence.
If any of them aren't true, you don't know yet. You have a vibe. And pivoting on a vibe just resets the clock on a channel that might have worked if you'd stayed with it another month.
Almost every founder who tells me "cold email doesn't work for us" turns out to have never done more than 300 controlled sends. The channel didn't fail. They abandoned it before it could show them a result.
The diagnostic question to ask yourself today
Two questions.
What is your current control message, and how many sends have you done on exactly that message, to your current ICP, in the last 60 days? If the answer is under 500 and you're already thinking about switching channels, you're about to make the same mistake most founders make. You don't have data. You have anxiety about the data.
How many variables have you changed in the last three weeks? If the answer is more than one at a time, you're running experiments in parallel and won't be able to distinguish what worked from what didn't. The fix is to pick the most important variable to test, hold everything else constant, and run 500 sends before touching anything else.
The founders who succeed at outbound aren't the ones with a special message or a magic channel. They're the ones who ran the same channel long enough to know what worked on it. This is the least glamorous of the four bottlenecks I see in early-stage SaaS sales motions and probably the easiest to fix. The framework isn't complicated. It's just patience with a stopwatch and a spreadsheet.
Fix the volume before you diversify the channels. You almost never need a new channel. You need the same channel, run 400 more times, with the same message.
