Looking for a market - Notes (Part 2)

Activation speed as a filter

Getting straight into it, this post is more about the things I’ve learned recently.

Building ProposalKit, I am seeing that it’s important to consider the speed and ease with which the customer becomes activated in your product. Because you can do all the tips and tricks in the book, but the customer must receive value pretty quickly out of your thing, especially if you’re not already an established brand with a lot of trust behind you. And you kind of need all the help you can get when starting out. For more involved use cases, sure you can make the activation paths longer, but not for the MVP.

It could be somewhat hard to understand that lesson before you’ve really thought through the main value drivers of the product you’re building. I also grasped it only once I understood that I cannot reasonably write the proposal for the customer. It’s the customer that understands the discovery call they just had, and what goes in this offer. We can only help with it, but it’s not meant to be a thing you do in seconds.

Read the market before anything else

Before anything else, I should have gone through acquire.com and the public revenue trackers like TrustMRR first, to look at the businesses that are making money in this space right now and, more importantly, growing. It’s easy to wave it away with “yes, of course the incumbents in the chosen space are making money, if they’re still there and making moves.” But really, I have no idea how much they make, whether they’re struggling, or whether there’s any meaningful problem left unsolved in that space.

On these sites, if someone makes over $100k a year, and that revenue is growing, plus there are other players in this space, this just might be a wave worth riding.

Mining reviews for a real wedge

I did go through G2 for reviews, but I went in by misusing the AI a bit. The search was already biased towards being able to build something beautiful, and that bias leaked into how it read the reviews. I should have lost that wording from all of its system contexts beforehand, as this was just the preselection criterion for my ideas, not something that should go through the whole pipeline. Probably, the only thing worth doing on G2 is to list and document every review that points at something specific and solvable. After that, I’d go over every review with my own eyes and cull all the ones that I know would mislead me and the AI systems. Then the test is simple: is this wedge sharp enough that I can go back to the exact person who wrote the review and say, “that problem you surfaced is now fixed, here, use it in my product”? If the complaint is specific, I can do that. If it’s vague like “the PDFs are sometimes broken and that’s annoying”, then the best I can come back with is “my PDFs are probably not as broken.” That’s a weak argument to build a foundation on.

The more important note: I didn’t pick the wrong idea per se. I went in thinking it doesn’t really matter which business I build, as long as I can actually validate it. So I deliberately picked at the cross-section of two things I cared about: something beautiful to work on, and something close to money. I might well pick that way again. What actually went wrong is that I let that beauty point run too heavy in the next stage. The AI tooling and my research came out far too primed on that one wording, that one idea, when I should have cut the beauty criterion off completely before moving from picking to validating. It’s an OK way to choose what to build, but it has no business steering the reviews search.

Adding a cold outreach stage

Now that I’m more familiar with Apollo, I think a proper validation should probably include a cold outreach stage too. I’m not 100% on this, but the targetability there is better than anything else I’ve seen. The structure would be: send out N emails, built around a problem statement I genuinely believe is strong enough, and let the responses decide. If people care, I continue. If nobody cares, I let the idea rest. And that’s fine, because any idea far enough from this one (the one under the magnifying glass), even after a few pivots, is technically a different business anyway.

The part I haven’t worked out is how to de-risk the message itself. If my message is bad, I can’t tell a dead idea apart from a badly-pitched one. With enough volume I could possibly do it in segments though, and that might do the de-risking enough. But it also means that it cannot be variations of the same message. Each segment should have significant differences, e.g., a different message per use case you’re trying to validate in that same market.

What makes a wedge “big enough”

And I’m still not sure what counts as a “big enough” wedge. I wish I knew better. The pain has to be strong and recurring, not mild and occasional. I should be able to name the switch trigger, the actual reason someone abandons a tool that already works otherwise. It should be a beachhead, which means that it’s narrow enough to completely win one segment, with an obvious next step once I have it. And it has to be falsifiable: I should be able to say in advance what evidence would prove it isn’t big enough.

With this chain of reasoning though, I’m afraid this kind of wedge might not actually exist, and I would burn out looking for it. But, naturally, no point in aiming low out of fear.

A more precise approach

The bigger change for me, going forward, is that I’m dropping the “it doesn’t really matter what I build, as long as I can validate it” stance. It gave me too much leeway. I’d rather give myself the best chance up front by picking more blindly, by the metrics, before taste or anything else gets a say. Concretely: start from spaces where I can already see comparable businesses doing real numbers, that is, something north of $100k ARR and still growing on acquire.com or the public trackers, and pick what’s very likely to work first. Whatever room is left after that filter is where beauty gets to choose.

Why growth matters

And one thing I want to note down before it slips, because I’ve unfortunately never put weight on it before: the growth part of that bar matters as much as the size. Well, I actively disregarded that point because I thought it was bullshit. But it’s not. A growing market means the industry isn’t stale, and a stale one is where “best” already exists, entrenched, with nowhere left for me to wedge in. If it’s still growing, then “best” hasn’t been settled yet, which means it can still be me, or at least I can make a real effort at it and take some part in the profits.

Sidenote: the growth I’m after might not live in the headline category. “Proposal software” as an industry could be flat while all the real growth is in something like AI-first proposal generators. So the job is to find where the growth actually lives, and aim there.

Counterpoint: some industries are old and still full of money, like finance, and the gambling industry. And another “cool-bank” or “cool-bet” kind of company will still make truckloads of money, because the industry is just big enough for this. But it probably also means that there are unwritten rules, and oddities that I don’t know before getting into it.

Current situation

These are just notes on what I’ve learned from launching ProposalKit, and what I’d do differently next time around. ProposalKit’s still testing Google Ads, and experimenting with cold emails from Apollo, but more manually right now. There are likely to be big changes in Part 3.