How We're Positioning a SaaS

Introduction

This is How We Grow, our weekly log of what we're doing to grow Kai. We skipped a month while the team was heads-down, so this one picks up a summer's worth of work.

Kai has been a product since March. We built it, named it, branded it, and kept iterating. It works: people use it every day, we get feedback constantly, and some of them pay. The hard part now isn't building. It's distribution, and the first question distribution asks is one we can't answer yet.

So I put it to Jim, who has been on this with David for two or three months: can we say today who we sell Kai to?

His answer was no. Not yet. Hypotheses and bets, another two or three months before we land it. This issue is what sits behind that no.

Prefer to watch or listen? Here's the full episode

Positioning is not finding customers

I asked Jim the question I suspect a lot of people have and don't ask: what's the difference between finding the people you want to sell to, and positioning?

His answer is that finding the people is the easy half, and it comes second. Positioning is understanding which group of people are trying to solve which job, what they'd use instead of you, and why you're different in their eyes. Once you have that, finding them is straightforward. Without it, you're guessing at every ad, every landing page and every sales call.

His example: say Kai is an executive assistant. Pitch that to a secretary and the first thing they think is "how is that different from what I already do when I forward an email to the right person". They're already comparing you against something. The comparison happens whether you chose it or not.

The version that stuck with me is one I stumbled into on the episode. You can't sell Moët & Chandon at Lidl. Not because the bottle is wrong, but because nobody walks into that shop looking for it. Jim took it further, and this is the part worth keeping. Put the same bottle in a supermarket's alcohol aisle and it competes with other drinks. Put it in the luxury lane next to a Louis Vuitton bag and it isn't a drink any more, it's a luxury good. Same bottle, different frame of reference, different price.

That maps onto us directly. Today Kai is on the AI assistant shelf and our plan is 50 dollars a month. Fyxer, on that same shelf, is 20. Someone comparing the two on price has an obvious answer, and it isn't us. The shelf sets the ceiling before anyone looks at the product.

The six steps Jim ran

Jim went and found a framework rather than arguing from taste, which I think is the useful part. He used April Dunford's cascade, from Obviously Awesome, and ran it per group of users.

The positioning cascade, applied to Kai.

Each step narrows. You start by clustering the users who love your product by what they actually do with it, not by who they say they are. Then, for that cluster, you list what they'd realistically do without you. Jim's version of this for the email crowd: connect a Gmail MCP to Claude, paste prompts into ChatGPT by hand, use Superhuman, use Fyxer, hire an actual assistant, or do nothing and keep writing their own emails. Doing nothing is always on the list, and for most products it's the one that wins.

From there you isolate what only you can do against that specific set of alternatives, translate it into why those people should care in their words, work out which of them cares most, and only at the very end pick the shelf. The order matters. Choosing your shelf before you know who cares means choosing your competitors at random.

The step that surprised me is the first one, because it means you can't do this before you have users. Jim was blunt about it: don't position a product before you've shipped anything and heard back from the market. You need a bet to start, then evidence to sharpen it.

I asked about the obvious counterexample. ChatGPT didn't do this. Jim's answer is that a genuinely new product can launch without a position and work it out afterwards, but that's rare enough not to plan around. He gave a better one from a conversation the team had recently with Anthony Pierri: when Apple launched the iPhone they could have sold it as a computer small enough for your pocket, which would have put it next to machines with keyboards and made it look absurd. They sold it as a phone that also does what a computer does. Same device, and the choice of shelf is most of why it worked.

Three groups, and the one we cut

Running the cascade with agents in the loop turned up three groups of top users, and they're genuinely different people.

The first delegates: email, calendar, meetings, hand it to Kai and get the time back. The second is deep in AI already, running long automated workflows, and uses Kai as the real-world touchpoint that feeds those workflows daily context from meetings, mail and calendar. The third treats Kai as a productivity copilot: plan my day, block focus time, tell me what I did and didn't do this week.

We can only launch on one. So the interesting decision is the one Jim made without running a test.

The third group is cut. Not because those users aren't real, but because of where they'd put us. Someone shopping for a productivity copilot compares Kai to Morgen, which is our own other product, and then to Sunsama and Routine. Those sit at 8 to 12 dollars a month. Our plan is 50. On that shelf we can't justify the price, willingness to pay is lower, and the audience skews toward students. It's the Lidl problem exactly, and no landing page fixes it.

That's a positioning decision made on the comparison set alone, before any money went into testing it. It's the cheapest kind of decision available and we don't make enough of them.

Testing the two survivors at once

For the two that survived, we didn't pick. We're running both.

Three Google Search campaigns, one per positioning, each buying keywords specific to that group and landing on its own page. What we watch is conversion on the page, what a signup costs, and whether those people stay in the app afterwards.

Then the creator pods, which started the day before we recorded.

How a pod works.

A pod is 10 to 15 creators on Instagram and TikTok who all reach the same kind of audience, all briefed on the same message, all posting in the same window. One pod pitches Kai as the executive assistant that runs your calendar. Another pod, picked from creators deep in AI, pitches it as the live context layer for your agents. Same product, two stories, two audiences, and the traffic each one sends tells us something the other doesn't.

The first collab went out and did more for us than the positioning read. We went from around 100 visits a day to over 1,000 in 24 hours. At 100 a day you can't test a homepage. At 1,000 you can, and we've started running angles against each other on the homepage and the onboarding because the traffic finally supports it.

The messier parts

Two of these bets are wrong, and we're paying for both. Jim said it plainly: if the three groups were picked correctly, only one of them is right, which means the money and the months going into the other two buy an answer and nothing else. He thinks it's worth it. It's still the most expensive line in this plan.

A bad video looks exactly like a bad positioning. This is the risk I pushed on. If a creator makes something weak, the numbers come back low and we can't tell whether the message failed or the execution did. Jim's answer is volume: with 10 to 15 creators in a pod, one weak post gets outvoted by the rest. It's a real answer and it isn't a complete one. The same problem exists on the ads side, where a poor landing page can sink a good position just as easily.

Nothing has produced a verdict yet. This is worth being clear about. What we have today is machinery running and one decision not to spend. Not results. Jim expects two to three months before the data says anything, and we're not going to pretend otherwise in the meantime.

We've already made one call on gut feeling. The homepage runs the AI executive assistant angle. Not because a test said so, but because that's what Jim and David believe Kai is. The other angle lives on its own landing page with campaigns pointed at it. His recommendation to anyone doing this is the same shape: pick the one you believe in, put it on the homepage, and test the others adjacently. Running two or three homepages at once gets out of hand fast.

What's next

September is the noisy month. More creators posting, more pods, and the traffic that comes with them.

Outbound, which is new for us. Jim and David have started getting people on calls and prospecting directly. That's a different motion from anything we've run so far.

A launch video. That one's mine, and it lands in October.

And we're hiring. We're looking for someone to run content with us in Lisbon. If that's you, get in touch.

The numbers, so this issue has some: 150 to 200 active users, about 15 signups a day, and 12 people paying. Jim's framing is that we're still pre-revenue, with a handful of people crazy enough to buy a subscription from a product this young. In two or three months we should be able to tell you which of the two positions those people came from.

If you want to follow along, the whole series lives at How We Grow, and the changelog is the receipts of what we ship week to week. Sign up and each issue lands in your inbox, along with early access to what we're building.