I’ve been running a technical writing agency, Frindle, for the past four years, and have worked with over 50 startups, mostly building developer-facing products.

I get asked all the time now how my business hasn’t been completely taken over by AI, but the reality is that I’ve never had a client churn because they have automated what my agency does, while I’ve had lots churn because they came in without a clear plan of what they wanted to achieve with the content/ended up pivoting/it wasn’t ultimately the right time.

So this is my full write-up of how founders should think about content for early to mid-stage startups, particularly B2B and technical ones.

Setting the right expectations for content

For the most part, content is a long-term marketing play. It’s for establishing your team as domain experts, building your brand, and building out your top-of-funnel growth.

It’s NOT for immediate sales.

In my experience with clients, conversion to signups, let alone paid signups, on any given article is low even for articles that e.g. go viral on Hacker News. If you're like, "I need this to get me new customers ASAP", it might not be the right strategy for you.

Two types of content

I typically think of content in two main pillars: brand content and direct response content. Brand content is just content to help people be aware that your company exists. This includes things like "state of MLOps", thought leadership pieces, etc.

Direct response content, on the other hand, is content that's supposed to be tactical for the audience. This is stuff like technical tutorials, how to accomplish any given thing with the software, anything SEO related.

Pillar 1: Brand content

Brand content is about producing content that’s going to do well on Twitter/LinkedIn and platforms like Hacker News. These should be long, insightful, in-depth pieces that showcase your domain expertise and demonstrate clear proof of work. (Proof of work is super important because with ChatGPT and Claude, it’s trivially easy to generate mediocre content and people’s attention spans are frayed. While AI can assist in content creation, your pieces all need to contain core nuggets of unique insight and knowledge).

Some examples include:

• Traditional engineering blog posts where you write about something interesting or hard that you built—the work itself serves as proof of expertise
• Vertical analysis pieces that bring unique expertise and insight to bear (think Stratechery-quality analysis)
• “State of” industry surveys where you had to do a ton of work to collect the data
• More “out-there” projects like Retool’s magazine-like piece on the history of Visual Basic or Modal Labs’ GPU Glossary or HuggingFace’s embeddings leaderboard.

Recommended Frequency: 1-2 pieces per month, consistently published and properly distributed, is sufficient. Quality over quantity is crucial here.

Pillar 2: Direct response content

Direct response content is meant to both provide immediate value to the reader, and also elicit an immediate response. The primary distribution for direct response content is search. So step number one here is understanding SEO, and making sure that your site is technically ready - that it’s fast, has a sitemap, etc, so that any subsequent pieces you publish rank well.

Step number two is understanding your Ideal Customer Profile (ICP) and what keywords they’re likely searching for. If you’re an analytics tool, for example, you might want to write something that will rank for “google analytics alternatives.” The key thing here is that you’re not just trying to shill your product. You want to actually provide a list of all the top Google analytics alternatives. You just happen to be one of them.

You should also think about what kinds of direct response content you can create at scale. Zapier is an example of a company that has done this well. They’ve programmatically created hundreds of tutorials for every possible combination of integrations within their 600+ integrated services. When someone searches "how to integrate X and Y," they're likely to land on a Zapier page.

Recommended Frequency: 2-3 pieces/week is enough to "keep things warm" for SEO. This should be very doable in a day's work—you don't need to produce thousands per year.

If everything is set up correctly, you should see results within a few months. If you're not seeing any increase in traffic after a few months, something is wrong with your setup, whether it’s keywords or site speed or something else.

(Mastering SEO, particularly in the AI age, is a whole other universe that I will talk about, perhaps in a future blog post.)

Third Pillar: Product Marketing Content

There's a third pillar, which is more traditional “product marketing”. I’m not going to talk too much about it here because this is well covered in other resources, but it encompasses things like:

• Documentation
• Technical guides
• Case studies
• Product updates

These are must-haves, and in fact, I would make sure these things are in place before moving on to the previous two pillars. This is content that helps educate people on how to use your product. Ensure that it is well-written, properly indexed, and easily searchable.

Tactical recommendations

Now that you know about the different content pillars and some basics on how to think about strategy, here are some tactical tips:

Use different content formats

When people talk about content, they usually think of written content. But text is only one of many formats that are open to you, and probably not even the most compelling one.

Every technical blog post should contain diagrams and cover art, and when you distribute on Twitter/LinkedIn you should include images and video.

You should also think about which format best suits a certain topic. For example, would a tutorial be better presented in video format? Or a demo as part of an interactive landing page? (Some examples here include Canva’s mp4 to gif page and ElevenLab’s text-to-speech page).

It is easier than ever to produce these sorts of “higher-effort” content with the help of Claude/ChatGPT.

Finally, re-use content! A video of a talk can be converted into a blog post, and a podcast into shorts. You’ve already done 80% of the work creating the initial piece of content, might as well get more bang for the buck with it.

Spend twice as much time on distribution as on writing

Creating great content isn't enough—you also need a solid distribution plan. You can't just write blog posts and not post them anywhere.

You don’t need to post all of these places, but you should pick a few that correspond to where your ICP/audience is likely to be hanging out online. And then you should invest in tailoring the blog post content to the particular platform. In particular, all the social platforms tend to strongly boost images and videos, so you should make sure to include those.

You should also spend some time thinking about who will post, and when they post.

In general, for early-stage startups, it’s more effective to post under the founder’s personal account rather than the company account. People are ultimately drawn to voices and characters, not corporate logos.

How to measure success

Views are the only reasonable metric to measure when it comes to content. Conversion from any piece of content, even viral ones, will likely be low — remember that content marketing is about creating many touchpoints with your brand and building trust over time.

Are you ready to invest in content?

If you are pre-product market fit, you’re better off focusing on outbound sales or VC introductions for lead generation. You should also spend time getting “the basics” in place like documentation and a few key case studies.

Post-market fit, though, and especially once you have a sense of who your ICP is, it makes sense to start investing in content. Start slow, maintain high quality, invest in the distribution, then ramp up on the channels that seem to be working for you.

Who to hire for this

I personally think that if you are building a technical product, you can get very far by hiring a technical person who likes writing and have them supervise ChatGPT.

Now, you’re only going to be able to get this kind of person if you’re willing to pay them on par with engineers, but over the long run, it’s still going to save you money because you won’t need to hire subcontractors/agencies.

However, if you can’t find someone like this, please feel free to reach out to my agency, Frindle!

Why am I giving away all this info for free?

You might be asking — why am I giving away all these insights for free?

And the answer is that content is all execution. It doesn’t matter if you know what to do, it’s still a ton of work to actually go do it!

If you are interested in having Frindle help you execute, please feel free to reach out!