Luke Whiting

2 hours ago

Everything We Tried Failed Except for This One Weird Trick

The other day Jamie and I were scheming about how to take over the world.

“It’s simple”, Jamie said as he stepped up to the whiteboard (the man loves an understatement). “People just have to continue to believe these three things about us, for a very long time:

1. I can trust Convex with important things.
2. Convex is alive, with undeniable product and ecosystem momentum.
3. Builders whose judgment I trust enthusiastically recommend it.”

So far, we’ve done a decent job of expressing these principles in the product. We’re not perfect (”Convex is a pretty good implementation of a great idea”, also Jamie). But people are building significant companies on Convex now; the love is there, too. We now are privileged to solve even thornier technical challenges for our customers (this is particularly exciting to James; I wish I could impress upon you the giddiness this man possesses for the finer points of scale-out infrastructure). The trust → aliveness → recommendation flywheel is showing sparks of success.

Let’s Go To Market

But Convex is not merely a beautifully rendered argument for how a complex piece of technology can and should feel; we’re also building a business here, one that scales fast and makes good on the contract we have with our investors and team. Do these principles still apply to how we Go To Market?

First, an introduction: I joined Convex in February to help us figure out how to Get That Money. Since joining, we’ve run lots of GTM experiments: we’ve sponsored newsletters, paid creators to make videos, siphoned thousands of dollars into Google and Reddit ads, hired an expensive AEO agency that tried to convince us to hook up a slop cannon to our website, and made a launch video that wasn’t quite right (watch it here and see if you agree).

Then we looked at the data. Nothing was working. Revenue growth was accelerating, but this appeared to be completely unrelated to any of our go-to-market efforts. Paid ads bought low-quality traffic (3.3x shorter average session times and 2.5x the bounce rate) and zero conversions. ~$100k on newsletters and creator sponsorships in two months brought a trickle of extremely expensive new visitors who didn’t actually build anything (some placements ran over $100 per click). And we’re deeply skeptical a slop cannon of LLM bait would move any needles we care about (the LLMs will win the AEO wars, not the agencies; we fired them).

Nothing, that is, except for our startup program.

Nothing works, except the thing that is expensive and hard to fake

The search for a fulcrum to help us bend our growth curve led us to the small-but-mighty cohort of customers who’ve participated in our startup program: turns out high performance, high growth startups, founded by people we respect do extremely well on Convex. On average, this cohort compounds their Convex spend 30x over 24 months, inclusive of credits. We empower their teams to focus on shipping cool stuff instead of futzing with Postgres connection pooling; in exchange they push us hard, which forces us to stay ten steps ahead of their scaling challenges. Convex gets better, the startups are (mostly) happy; they then recommend us to their friends. The virtuous cycle continues.

It’s worth noting here that the bar for foundational infra providers is (and should be)…high. For most companies, but especially startups, choosing the right infrastructure provider is an extraordinarily risky decision. Winning customer trust is a brutal slog (and harder to hold together than plasma); this is why conventional wisdom says you should build on a boring, battle-tested stack with deployment counts in the trillions and save your innovation tokens for the core product. You should most definitely not build your company on some new-fangled backend run by a tiny startup, no matter how smart you think the founders are.

And yet increasingly, they are.

Why?

Many of these founders have known Jamie and James for years, or worked with them at previous companies – privy to a rich source of longitudinal performance data on the subject(s) (at this stage, startup culture is a mirror of founder personalities and values, for better or worse). Some found us through organic recommendations from influential voices like Theo or other trusted friends. It helps, too, that both have been working on hard infrastructure problems, publicly, for decades prior to Convex (see the Magic Pocket project at Dropbox or one of James’ many published papers on error reduction in distributed systems, like this one). These signals are expensive and hard to fake.

That, and the product happens to be pretty fucking good, even to a recovering vibe coder like me.

Whole ass one thing, never half ass two things

Ron Swanson understood that effort has a convex (heh) output curve: the opportunity cost of not focusing on making the thing that is already working work even better can be catastrophic. In the age of slop-cannon-all-the-things, when everything can be deepfaked, we expect the return on that which cannot to go vertical in the coming years. (Probably why Joshua Kushner and Bob Iger bought the Lakers for a record-breaking valuation.)

So we’re shutting down the failed GTM experiments, and de-investing in shoddy signals. Instead, we’re tripling down on what’s already working. More high-quality in-person events; less feeding the ad targeting leviathan. Larger budgets for supporting the customers that push us hardest; smaller budgets spent buying clicks that bounce. More earnestmaxxing and unfiltered poasting, fewer paid sponsorships. We’re going to make it easy to get to know us deeply, to create the conditions in which your trust can flourish; organically, unforcedly.

This is going to take some experimentation and time; we’re still finding our voice. But with any luck, our conversation with Mr. Market will start to feel like a party you never want to leave (or, for the introverts amongst us, a project you never want to finish).

If you’re a startup building something serious and we sound like your kind of infra partner, you can apply to the startup program here.