A bird alarm I never bet on quietly beat every idea I loved, and figuring out why meant admitting I’d been reading my numbers backwards for years.
I build things for fun, and to fix my own problems. That’s pretty much the whole business plan. There are no VCs involved anymore and no growth team waiting on a standup, it’s me, Leandra, some friends giving me priceless nudges, a credit card with a nervous limit, and a list of small annoyances I’d like to make smaller.
You might assume the part I obsess over is the building, but it really isn’t. I love building, it’s my comfort zone. The thing that keeps me up at night is the marketing experiment. I have a tight budget and even less time, and I have to decide where to spend both. Get that wrong and the nicest app in the world dies quietly in the App Store, unranked and ignored.
So this is a story about getting it wrong on purpose, cheaply, and about what a bird I didn’t even fancy the odds of ended up teaching me.
The control group won
I’ve already told the long version of this story over here, so here’s the short one. Before building anything, we ran Facebook ads to test a handful of alarm ideas against each other: Social Alarm (wake up to secret messages from people you love), an alarm from your future self (my real darling), one with quotes, and one with a sexy voice, all of them against the boring baseline I’d thrown in only so the others would have something easy to beat, an alarm with birds.
The birds won, and not by a hair. I’d rooted for Future Self right to the bitter end, and it didn’t matter at all, because the market doesn’t care who I’m rooting for. A cheap little experiment had just told me, very politely, that my taste was the least valuable thing in the room.
Ahh, but your gut is special, Dieguito, no?
– My Favorite VC
It isn’t, and the dashboard had just said as much out loud. What I didn’t understand yet was why the bird kept winning, and that took a second app to figure out.
So I ran the same experiments on two very different apps
I don’t have just one app. I built more than 10, and most of them are gone or don’t get much of my attention, but one is still a good runner, Smart Keys (a smart keyboard for people like me who write in a second language and would rather not sound awkward), and now there’s Bird Rise (the alarm that wakes you up gently with birdsong). They couldn’t be more different from one another. One solves a writing problem and the other solves a waking-up problem (waking-up problem??? Really??), and the people who use each one barely overlap. Smart Keys lives in the middle of your sentences, while Bird Rise owns your first waking minute.
The same broke person built both of them, on one tiny budget, and ran them through the same playbook, even though the apps themselves share almost nothing. And that, it turns out, is the most useful science I’ve ever run on myself.
Experiment one: buying users
The first experiment is the boring and expensive one. You pour money into Meta, watch who comes out the other side as a paying customer, and then do the math on CAC against LTV (what it costs to win someone, set against what they end up paying you back).
Smart Keys barely scrapes an LTV/CAC of 2. It works, but only just, and you can almost feel it sweating to get there. And that was fine, my strategy is to have many of those little apps with tight margins that together could make a difference. But then, Bird Rise tops 3 and keeps climbing. (the holy LTV/CAC > 3 that VCs supposedly swoon over, that one??)
The download-to-paid story is even more lopsided. Smart Keys turns roughly 3% of downloads into paying users, while Bird Rise sits comfortably north of 6%. That’s two times the conversion out of the same channel, produced by the same clueless guy.
I’ve dropped over 30 grand (most of it straight into Zuck’s pocket) to learn things like this, so here’s the receipt for free. Two apps are not equally good at turning attention into money, and the gap has almost nothing to do with how hard I worked on either one.
Somewhere along the way I started thinking about these apps less like products and more like a portfolio I angel-invest in, each one a little bet with its own risk profile (though plenty of them I still build purely for fun, or for a cause, and those just pay out in a currency these charts can’t see).
Which raises the uncomfortable question any honest indie dev should sit with: would all this money have done better quietly parked in a boring index fund? I once built myself a chart putting my profits next to my ad spend next to what a conservative investment would have returned over the same years, and it was not a flattering picture (a reckoning I’ll save for its own post).
Experiment two: the hard paywall I didn’t believe in
Then I read a RevenueCat State of Subscription report claiming that hard paywalls (the kind where you hit a wall and have to pay or start a trial before you really get going) can actually work, even for small apps.
I didn’t buy it. A hard paywall felt like the sort of move only a brand with real trust and real money could pull off, and we have neither of those. It seemed obvious that a tiny app run from our apartment couldn’t slam the door in a stranger’s face and ask for money (or commitment before testing the product).
So I ran the same hard paywall A/B on both apps at once. On Smart Keys it was a flat, immediate no. On Bird Rise the test is still running, but the early read (which I’ll trust before the final number lands, because the direction is already that clear) is that the hard paywall might actually work, on my little app with no brand to lean on, which I honestly didn’t see coming.
I didn’t believe the RC report, and the report didn’t care. It was the same experiment producing opposite answers, which is the whole point. There’s no playbook you can lift off a blog post (this one very much included), only the version you test on your own thing.
Weren’t you the king of A/B tests, Dieguito?
– VC-backed Me
Prince at best. In a past life I was a properly funded tech bro out to change the world from Silicon Valley, and I ran dozens of these tests there, burning through millions of dollars in the process. The punchline is that when we finally shipped the winning variants, not a single metric that mattered actually moved. I came out of all that pretty burnt out on the whole A/B religion, so this (cheap little tests on my own apps, with my own nervous credit card, where the answers actually change what I do next) is me quietly making peace with it. And even half-converted, I learned more from this one cross-app comparison than from a hundred button-color experiments back then.
What kept showing up
So why does the bird keep winning experiments I never designed it to win?
It took me embarrassingly long, and a lot of squinting at Mixpanel, to notice it. Every metric where Bird Rise pulls ahead (it’s cheaper to acquire users, and they convert and commit far more readily) traces back to something the flashier ideas never really had: a strong tribe.
“Tribe” is a slippery word, though, so let me pin down what I mean by it, because it isn’t a big audience and it isn’t a subreddit that happens to exist. It’s the difference between people who merely share an attribute and people who’d put that attribute on a mug. Here’s the test I use:
| A tribe | Not a tribe (just a category) |
|---|---|
| They wear the name with pride: “I’m a birder” goes straight in the bio. | A label in a demographic report. They don’t proudly tell anyone. |
| They could talk about it for 24 hours straight. | They’ve got nothing in particular to say about it. |
| There’s a mug, a tee, a sticker, and they own it (or want it). | No one would ever put it on a mug. |
| It comes with its own lingo, rituals, heroes and rules. | It’s just an attribute people happen to share. |
| Members quietly rank each other by how deep they go. | There are no insiders, no status, no ladder to climb. |
| They evangelize, dragging friends in (like: vegans, fasters, CrossFitters). | There’s no reason to tell a soul. |
Still, “Tribe” isn’t a yes/no, it’s a gradient.
Birders immerse themselves fully (with their lingo, gear, life lists, and even a century of Audubon membership behind them). People learning English as a second language form a real community, but it’s a gentler one, more a network of practical groups than a single passionate culture. Within this, specific language-learning groups (Korean learners, Portuguese from Portugal versus Brazilian, Duolingo’s owl fans) are noticeably stronger than the broader community. Smart Keys benefits from these smaller groups but never quite achieves the strong pull that birders give Bird Rise, which accounts for most of the difference between converting at 3% and 6%. Future Self had neither of these, just my own private fantasy with an alarm attached, while the bird came with a whole community already waiting at the door.
The thing sitting underneath all of it is a daily ritual (a natural one like waking up, not a streak you guilt yourself into) resting on top of a community that already cares, something I’ve started thinking of as Ritual × Tribe. That mix is cheaper to reach and far more willing to commit. I can’t prove it’s the strong combo and not just dumb luck, but every cut of the data keeps pointing the same way. It’s not that I magically turn into a better builder when I build for birders, it’s that I happened to pick a better pond, one that came stocked with both halves (entirely by accident), and the data spent a whole year patiently trying to point that out to me.
And here’s the part that took me a while to truly understand: this chart isn’t a scoreboard. Every corner of it has winners. A useful, tribeless app can definitely generate money if you can buy the channel or own it outright. A big, passionate crowd without a daily hook can succeed through merch, sponsorships, and partnerships. None of those are bad quadrants; they require different marketing strategies, and each one runs on resources I don’t have: funding, a sales team (or any team at all), a channel I control, or a B2B approach.
This chart has a couple of other dimensional axes I’d love to explore more, but let’s keep it simple for now so I can convince myself that I created a model to evaluate my ideas.
Smart Keys and Bird Rise accidentally dragged me to the top-right quadrant. It’s the one corner I’ll aim for on purpose from now on, not because the others lose, but because it fits my reality (budget, team, expertise, lifestyle). Until I get excited by some cause or have fun with some technology that makes me forget about all of this. ;p
Metrics to change, or metrics that change me?
Underneath the bird and the tribe, there’s a bigger shift this whole thing forced on me. When I see a download-to-paid number stuck at 3%, my trained instinct is to treat it as a dial: tighten the onboarding or move the price around, and try to drag that number up by force. That’s the optimizer in me, the part Silicon Valley trained into me and paid handsomely for. But a metric can be a clue instead of a dial. Bird Rise converting at two times the rate of Smart Keys was never really a paywall problem I could fix, it was the market quietly telling me something true about the concept itself, about whether the need ran deep and the people were hungry enough.
The only way I’ve found to tell a dial from a clue is to bet one idea against another. You don’t learn the shape of something by staring at it, but by holding it up next to whatever it isn’t. A number on its own just begs to be optimized, but the same number sitting next to a wildly different app stops being something I turn and becomes a verdict on my thesis, on what I’d assumed about the problem and whether anyone needed the product as badly as I did.
I guess some metrics are there for me to change, and the more important ones are there to change me.
