Almost everything written about marketing assumes you have a thing that sells and need more people to buy it. Startups before product-market fit have a different problem: you don’t yet know exactly who buys, why they buy, or what words make them buy, and no amount of reach fixes not knowing that. Survey data suggests only around half of startups have anyone dedicated to marketing at all; it’s the founder, at 11pm, boosting a post and hoping.
So here is the honest playbook for that stage, from an agency that regularly tells startups at this stage not to hire us yet. Pre-PMF marketing is not a growth machine. It’s a learning machine that occasionally produces customers as a byproduct. Build it that way and the growth machine comes later, cheaper.
Step 1: An ICP narrow enough to embarrass you
Every founder claims to have an ideal customer profile. Most have a demographic horoscope. The test is whether your ICP excludes people you’d happily take money from. If it doesn’t, it isn’t narrow enough to teach you anything.
Fake narrow: “small business owners who want to grow.” Actually narrow: “independent physiotherapy clinic owners with 2 to 5 staff who are losing bookings to chains and currently manage everything through a receptionist and a paper diary.” The second version tells you where they gather, what words they use, what they’ve already tried, and what to say. The first version tells you to buy broad ads and pray.
Narrow is a starting wedge, not a prison sentence. You expand after you’ve won somewhere. Every recognizable brand you admire started embarrassingly specific and grew out of it; you’re just seeing the after photo.
Step 2: Spend founder time before ad money
Early on, your scarcest marketing asset isn’t budget; it’s the credibility and knowledge in the founder’s head. Early buyers trust a person with a point of view long before they trust a logo, which is why founder-led distribution (posting, showing up in the niche communities where your ICP already gathers, doing direct outreach that doesn’t smell like a sequence) reliably beats small ad budgets at this stage. It doesn’t scale, and that’s fine. You don’t need scale yet; you need conversations. We’ve written up the sustainable version of founder-led marketing separately, because “the founder should post more” is useless advice without a system.
One warning about the obvious shortcut: AI can draft your content faster than ever, and everyone else’s too. HubSpot’s State of Marketing research found the overwhelming majority of marketers now using AI for content, and over half saying the resulting saturation makes differentiation harder. The things a model cannot generate (your specific experience, your customers’ stories, your actual opinions) are the only inputs that were ever going to work at this stage anyway.
Step 3: One channel, ninety days, no cheating
The most common self-inflicted wound we see: a founder running thin efforts on five channels, learning nothing from any of them. Every channel needs a minimum viable dose before it tells you the truth (content compounds over months, outreach needs iteration on the message, communities need genuine participation before they tolerate a link), and five channels at one-fifth effort each delivers five inconclusive experiments.
Pick one primary channel based on where your narrow ICP actually discovers things, commit for ninety days, and define in advance what result would justify continuing. Not followers, not impressions: qualified conversations, signups, or revenue. At day ninety, double down or switch, with data either way. The discipline feels slow. It is dramatically faster than the alternative, which is being confused indefinitely.
Step 4: Let customers write your copy
Before product-market fit there is a smaller, earlier fit to find: message-market fit. The words. And the reliable shortcut is that your best copy already exists in your customers’ mouths. Interview the people who bought, the people who almost bought, and especially the people who left. The phrases they use to describe the pain are your headlines; the objections they raise are your FAQ; the alternatives they compare you to are your positioning. This one is so cheap and so consistently transformative that we gave it its own post.
Step 5: Measure learning, not applause
Vanity metrics are especially dangerous pre-PMF because they’re available before revenue is. Followers, impressions, and likes will grow with effort whether or not anyone will ever pay you, which makes them a comfort blanket, not a compass. The dashboard that matters at this stage fits on an index card:
- Qualified conversations per week, however they arrive.
- Where each one came from, tracked by asking, because your analytics can’t see a group-chat recommendation.
- Landing page conversion rate, because messaging problems show up here first.
- What you changed and what happened, one line per week. This is the learning log that becomes your marketing strategy, written by reality.
The don’t-do list
Equally important, the things that look like marketing and mostly burn runway at this stage:
- PR and polished brand campaigns. Awareness doesn’t convert when the audience doesn’t yet understand the category or feel the urgency. Press is a lovely amplifier for a machine that works; it is not the machine.
- Hiring a marketer to “figure it out.” If the ICP, message, and first channel are unvalidated, you’re handing a stranger a research project with a salary attached. Validate the motion, then hire someone to scale it. (Yes, this applies to hiring agencies too. Including this one. Our outgrown-DIY checklist is the honest test for when that changes.)
- Copying visible tactics from big companies. A Series C brand runs broad content and sponsorships on top of invisible prerequisites: existing demand, trust, data, and a team. Copying the visible half without the invisible half is cargo-cult marketing.
- Scaling ads before the page converts. Paid traffic into an unvalidated message is the fastest way to turn money into confusion.
What graduating looks like
You’ll know this phase is ending when the machine gets boring in the best way: the same narrow audience keeps converting, the same message keeps working, one channel produces predictably, and customers start using your words back at you. That’s the moment marketing stops being research and starts being an investment case, and it’s the point where budgets, hires, and yes, agencies, stop being premature and start compounding. If you want a sanity check on whether you’re there, that’s a conversation our strategy team has weekly, and we’ll tell you honestly if the answer is “not yet, and here’s what to learn first.” We have a well-documented habit of that.