Table of contents
In 2022 we tried to scale our content before we’d learned anything about it. We picked a keyword, fed a brief into an AI writing tool, and published the entire article without a human rewrite. Two months later it had zero clicks, sat at position 33 in Google, and seemed to drag the rest of the site down with it. The note I left on the campaign reads: “Don’t experiment with AI only content on growthmethod.com.”
It wasn’t a tooling problem. We’d skipped straight to the scalable version of something we had never done well by hand.
That mistake is what Paul Graham wrote about in Do Things That Don’t Scale. The essay is aimed at startup founders, but I’d argue in-house B2B marketers need it more. Founders at least know they’re at the start. Marketing teams inherit budgets, tool stacks and targets that assume they already know what works.
Graham wrote it for founders, but marketers need it more
Graham’s core argument is that startups don’t take off on their own. Someone has to push them, usually by hand, long before any system takes over. His most quoted line:
“You can’t wait for users to come to you. You have to go out and get them.”
Paul Graham, Do Things That Don’t Scale
Swap “users” for “pipeline” and you have most B2B marketing plans. We launch a webinar and wait for registrations. We publish a gated guide and wait for downloads. We switch on a nurture sequence and wait for sales-qualified leads. It all scales, and very little of it teaches us anything.
Every new marketing campaign is a small startup. A new audience, a new channel or a new message has the same problem a new company has: you don’t yet know what works. The difference is that marketers are handed the tools for scale on day one.
Why marketing teams scale too early
There are three forces pushing marketers towards scale before they’re ready.
The tools are already paid for. Your marketing automation platform, your ABM platform and your outbound sequencer all promise volume. Once the contract is signed, not using them feels like waste.
AI has made volume nearly free. Anyone can produce 50 blog posts, 500 personalised emails or 5,000 ad variants this afternoon. When output costs nothing, output stops being a signal of anything. Our 2022 test was an early version of this lesson, and it only gets sharper as the models improve.
Dashboards reward volume. Big numbers look good in a quarterly review. “We emailed 40,000 contacts” sounds better than “I spoke to 12 customers”, even when the 12 conversations are what tell you why the 40,000 emails aren’t converting.
The irony is that the growth marketers actually want doesn’t come from one big push. It comes from small gains that compound. Graham makes the point with users, but it works just as well for leads, pipeline or revenue:
“If you have 100 users, you need to get 10 more next week to grow 10% a week. And while 110 may not seem much better than 100, if you keep growing at 10% a week you’ll be surprised how big the numbers get. After a year you’ll have 14,000 users, and after 2 years you’ll have 2 million.”
Paul Graham, Do Things That Don’t Scale
Ten more customers next week is a very unscalable target. You can hit it by hand. And the learning you get from hitting it by hand is what makes week 30 possible.
Five ways to do things that don’t scale as a marketer
Graham lists several tactics in the essay. Here are five that translate directly to marketing, with examples from our own campaigns. Before you start, a warning from the essay: they won’t feel like progress.
“The right things often seem both laborious and inconsequential at the time.”
Paul Graham, Do Things That Don’t Scale
1. Recruit users manually and individually
In 2024 we wanted more customer conversations, so we took the scalable route twice. First, an on-site survey asking visitors what they came to do, running on the website and in the app for 28 days. It got zero responses. Then a widget offering customer interviews, with a dedicated booking page. It got nine clicks in 28 days. My verdict on the campaign: “not worth it.”
Both campaigns waited for people to come to us. The feedback that actually arrives comes from reaching out to one person at a time: a personal email after someone signs up, asking a single question and promising a reply.
For in-house marketers: before you build a nurture sequence, send the first 50 emails by hand and read every reply. Before you launch a customer advocacy programme, phone ten customers. We’ve written a full playbook for this in send 100 manual cold emails before buying any tools.
2. Take extraordinary measures to delight new users
Graham argues that conventions set the floor for customer experience, not the ceiling:
“Once you realize that existing conventions are not the upper bound on user experience, it’s interesting in a very pleasant way to think about how far you could go to delight your users.”
Paul Graham, Do Things That Don’t Scale
The B2B SaaS convention is a signup form, an automated welcome email and a product tour. We went the other way. Getting started with Growth Method is one 30-minute call with me, where we connect your analytics and launch your first agent live, on your data. It caps how many new customers we can take each week, and that’s deliberate. Every call shows us where people get stuck, what they expected, and which words they use to describe the problem.
For in-house marketers: pick one moment in the buyer’s journey and make it absurdly good for a small group. A personal video for every demo request this month. A handwritten note to every new customer this quarter. See what it changes before deciding whether it’s worth systematising.
3. Focus on a deliberately narrow market
Graham’s point is that it’s easier to catch fire in a small space. Facebook started with Harvard students before it was for everyone.
Growth Method made the same choice. Asana, ClickUp and Monday.com serve anyone who manages work. We only serve marketing teams, and we centre the product around your goal, singular. A narrower market means every campaign, article and onboarding call speaks to one type of person, which makes it much easier to learn what resonates.
For in-house marketers: your total addressable market isn’t your campaign audience. Run the next campaign for one industry, one job title, or one region. If the message lands there, widening it is easy. If it doesn’t land anywhere, a wider audience only hides the problem.
4. Pick a single user and act like a consultant for them
Graham suggests picking one user and building exactly what they need, as if you were their consultant. You learn their workflow in a depth that no survey can reach.
We’ve done this with Colt Technology Services. Instead of pointing them at documentation, we worked inside their marketing process, down to writing a reusable skill so the AI agents produce copy in Colt’s brand voice and house style. Much of what we learned there became product. Laura Perrott at Colt put the results like this:
“We are on-track to deliver a 43% increase in inbound leads this year. There is no doubt the adoption of Growth Method is the primary driver behind these results.”
Laura Perrott, Colt Technology Services
For in-house marketers: pick one sales rep, one key account or one product manager, and treat them as your only client for a month. Build the exact content, campaign or report they need. You’ll find out what the rest of the organisation needs too.
5. Solve the problem manually for as long as you can, then automate the bottlenecks
This is where marketing teams go wrong most often, and it’s the lesson from our 2022 content test. Automation should come after you’ve learned what good looks like, not instead of learning it.
Our SEO page refreshes followed that order. For a long time I did them by hand. In 2025 I rewrote our ICE framework article after it slipped in Google, and it went from position 3 back to position 1. In 2026 I rewrote the title tag on our rule of 7 article to improve click-through rate. Click-through rate doubled, but the average position fell from 3 to 8, because the new title no longer matched what people were searching for. I reverted it.
Both campaigns fed straight into the instructions for our SEO agent, including a new rule to match the title to the top query driving impressions. Now an automated agent proposes and runs these refreshes on a schedule, and I review the results. It can only do the job well because I did it badly by hand first.
For in-house marketers: write down the steps every time you run a campaign manually. When the steps stop changing, the results are consistent, and the manual work is what stops you from running more campaigns, you’ve found the bottleneck to automate. If you can’t explain the rules to a new hire, you can’t explain them to software or an AI agent either.
How to know it’s time to scale
Doing things that don’t scale isn’t a permanent state. It’s how you work out what deserves to scale. Three signals it’s time to automate:
- You can write the playbook. The steps, the edge cases and what “good” looks like are all written down.
- The results repeat. The campaign has worked more than once, for more than one segment.
- You’re the bottleneck. The manual work is the reason you can’t run more campaigns, not the reason they succeed.
If only one or two of these are true, keep doing it by hand.
Recap: what to consider as a marketer
Before you buy the next tool or turn on the next sequence, ask whether you’re scaling something you’ve actually learned. Then work through the five:
- Recruit users manually and individually. Don’t wait for forms, surveys and gated content to fill your pipeline. Go out and get conversations one at a time.
- Take extraordinary measures to delight new users. Treat B2B conventions as the floor. Make one moment exceptional for a small group.
- Focus on a deliberately narrow market. Win one segment before you widen the audience.
- Pick a single user and act like a consultant for them. One sales rep, one account or one stakeholder, served in depth.
- Solve the problem manually for as long as you can, and then gradually automate the bottlenecks. Automate what you’ve learned, never what you’re hoping works.
Treat the unscalable work as campaigns
The hard part of doing things that don’t scale is that the learning gets lost. Ten customer calls live in someone’s notebook. The manual email test lives in someone’s sent folder. When it’s time to automate, nobody can remember what worked.
In Growth Method, every piece of unscalable work is a campaign that follows the same plan-launch-analyse workflow as your biggest launches. You plan it with a clear hypothesis, launch it, analyse the results, and record what you learned. When a manual campaign has worked enough times to become a playbook, the same record becomes the instructions for an AI agent to run it. Plan, launch, analyse, repeat, and scale only what the analysis says is worth scaling.
Get started to see how it works on your own campaigns.