You can correctly identify 3 growth problems and still fix the wrong one
I've made this mistake more than once.
You open the numbers, find three ugly things, and suddenly you have a roadmap.
Traffic is down.
Conversion is weak.
Churn is higher than you'd like.
All three can be completely true.
The problem is that "this metric is bad" and "this is what is stopping us from reaching the goal" are two different claims.
That's the part I think we skip too often.
Say the goal is more revenue.
Low traffic might matter. But before deciding to fix traffic, I'd want to know who isn't arriving.
Maybe the customers with the strongest reason to buy are still coming in just fine.
Conversion might be weak too. But weak for whom?
If one type of customer converts well and another barely converts at all, improving the average conversion rate is already a very different problem.
Same with churn.
Is the product failing to retain good-fit customers, or did we acquire people who were never a great fit in the first place?
This is why I've been trying to work backwards from the goal instead of forwards from whatever number looks worst.
What has to happen for that goal to move?
For which customer?
What are they actually trying to get done?
What path do they take to get there?
Where does that path break?
Only then does "what should we fix?" start becoming a useful question.
I'm curious how other people make that distinction.
When you find several real problems in the same business, what convinces you that one of them is actually causing the outcome you care about?
The "bad for whom" reframe is the part I keep having to relearn. It's so easy to treat a low conversion rate as one problem when it's really two or three different segments averaged into a number that describes nobody.
For your question at the end: what usually convinces me one problem is the actual cause is whether fixing it in my head changes the outcome downstream. Like, if I imagine traffic magically doubling, does revenue actually move, or do those extra people just bounce at the same weak conversion point? Running that little thought experiment on each problem usually exposes which one is the real bottleneck and which ones are just symptoms of it.
The other tell for me is talking to the people who almost bought and didn't. They point at the real constraint way faster than the dashboard does. Numbers tell you something's broken, the conversations tell you why.
I really like the “magically fix it” test.
The part I’m still trying to separate is what happens after a problem passes that test.
Say doubling traffic would clearly increase revenue. That tells me traffic actually matters and isn’t just an ugly number on the dashboard.
But conversion might pass the same test. Retention might too.
So I’m starting to think there are two different questions hiding here:
“Does this problem materially affect the outcome?”
and
“Of the problems that affect it, which one is actually worth working on next?”
Your point about talking to the people who almost bought feels especially useful because it gives you something the aggregate metric can’t: what was actually getting in the way of the decision.
How do you make that last jump when two problems both pass the counterfactual test? What makes one become the priority?
Ive done the exact thing you describe, open the dashboard, spot three ugly numbers and suddenly i feel productive cause now i have a to-do list. the reframe that got me was your for which customer bit, i realised my numbers were basically one blurry average hiding two totally different people. im early enough that my real problem is just nobody arrives yet so i cant even play the which-of-three game, but working backwards from the goal instead of forwards from the worst number already stopped me optimising a signup page that like 4 people ever saw lol. the thing that convinces me a problem is the real one is when i can trace an actual person to it, a churn reason, a why-i-didnt-buy. if its just a number moving i dont trust it yet. do you draw the line the same way or wait for more than one signal?
I think the “4 people saw the signup page” example gets very close to the distinction I’m trying to make.
The signup page might genuinely be bad.
You could even prove that its conversion rate is weak.
But fixing it perfectly still might do almost nothing for the business because hardly anyone reaches it.
On your question, I don’t think I have a fixed number of signals where I suddenly trust the diagnosis.
I’m starting to think about it more as: how much does this piece of evidence help me rule out the other explanations?
One person saying “I didn’t buy because of price” might only create a hypothesis.
But if their behavior is consistent with that explanation, similar lost customers say the same thing, and the alternatives start making less sense, confidence changes pretty quickly.
The opposite can happen too. You can have a lot of dashboard data and still not know whether the real issue is audience, value, trust or friction.
So I think I care less about the number of signals than whether independent evidence is converging on the same explanation.
What if a customer tells you one thing but the behavioral data seems to point somewhere else? Which one do you trust first?
That reframe – "does this evidence help me rule out the other explanations" – is a much better mental model than counting signals. I'm going to steal it.
On your last question, I actually hit this exact tension this week. People here engaged with my post, commented, seemed genuinely interested – the "customer telling you something" signal. But my actual subscriber count barely moved. If I'd trusted the stated interest, I'd have concluded the content or the pitch was working. The behavioral data said otherwise.
What resolved it for me wasn't picking one to trust over the other, it was noticing they were answering different questions. The comments were telling me the topic resonates. The lack of signups was telling me something about the path between "interested" and "subscribed" – friction, not disinterest. Once I stopped treating them as competing signals about the same thing, both became useful instead of contradictory.
So maybe the rule is: before trusting one over the other, check whether they're actually measuring the same decision.
Yeah, I think you just got to the part I find most useful.
Instead of treating “people engaged but didn’t subscribe” as one contradictory result, you can isolate the actual move:
And then the question changes completely.
What would make someone who’s already interested actually say yes to that next step? And what would make them stay where they are instead?
That’s why I’d be a little careful calling it friction already. Friction could be one reason to say no, but so could “I like this idea, I just don’t need more of it in my inbox.”
In your case, what do you think is the strongest reason someone who liked the post still wouldn’t subscribe?
Fair pushback — "friction" was me reaching for the comfortable answer, because it doesn't implicate the pitch itself.
Sitting with your actual question: I think the strongest honest answer is that "another newsletter about how people found customers" isn't differentiated enough at the moment someone's scrolling past a comment. The idea resonates in the abstract — people here clearly relate to the distribution problem — but resonating with a problem isn't the same as wanting a recurring email about it. There's no proof-of-value moment in a comment thread the way there is when someone reads an actual story and thinks "I want more of exactly this."
So the real test isn't engagement with the post, it's whether someone who reads one full case actually subscribes after that — a much narrower and more honest measure than "commented on my post." I haven't been tracking it that specifically. I probably should be.
Yeah, this is pretty much the idea behind Hacknator.
Instead of looking at the whole thing as “my post got engagement but no subscribers”, you narrowed it down to one specific movement: someone experiences the value, then decides whether they want more of it.
And I think your point about differentiation gets to the other half of it. Once you know which movement matters, the interesting question is why someone would make it or not.
In your case, “I liked this” clearly isn’t enough. “I want more of exactly this” might be.
That’s basically what I’ve been trying to build around: identify the movement first, then understand what makes people say yes or no to it.
That's a good place to land it. Curious how you're planning to measure "want more of exactly this" for Hacknator specifically – a repeat-usage signal, or something you'd just ask people directly?
I think repeat behavior is the stronger signal.
For Hacknator , the interesting moment isn't just whether the first diagnosis feels useful. It's whether someone comes back when they have another real growth decision to make.
I'd still ask directly, mostly to understand why they came back or didn't. But I wouldn't want a positive answer to substitute for the behavior.
That's exactly what I'm testing in the private beta now. Run your business through Hacknator and see whether, after the first diagnosis, it feels like something you'd come back to when the next growth decision shows up.
This resonates so much. It's incredibly easy to get pulled into fixing symptoms instead of diseases. What's been crucial for us is taking an almost forensic approach – instead of just seeing "low conversion," asking why specific segments aren't converting. Often, it's not a universal problem, but a mismatch for a particular type of user, or a specific friction point in their journey that's easily overlooked when looking at averages. Pinpointing that exact 'who' and 'where' the path breaks, as you put it, has been the real game-changer in finding what actually moves the needle.
Yeah, this is the part I’m curious about.
When you say you start asking why specific segments aren’t converting, how do you usually get to those segments in the first place?
Do you notice a pattern in the data first: source, region, behavior, device, whatever; And then investigate it? Or do the patterns usually come from conversations and qualitative stuff first?
for early-stage products, the patterns almost always come from conversations first. When you're just starting, you often don't have enough quantitative data for reliable patterns to emerge.
Instead, talking to your first few users (or even people who didn't sign up) from different channels gives you rich qualitative insights. You start hearing similar objections, confusions, or "aha!" moments that point to specific segments. Once you have those hypotheses from conversations - "people from forum X are struggling with Y feature" - then you can dive into what little data you have to see if it backs up those qualitative observations. It's about letting human stories guide your data investigation.