On Product Led Growth

2026.04.05

XCLVI

[From the Fall of the West to PLG; Curing Neuroticism; Knee Capping Asymmetry; PLG; Building The Mouse Trap; Grass is Greener; The AI Bubble Short]

Thesis: BirdDog is experimenting with Product Led Growth (PLG) to double down on our advantages and short the AI Bubble even more than we are.

[From the Fall of the West to PLG]

I know I’ve been riffing on how we should get together and save the west. To contrast that, today, we’re going to zoom in a little and talk more strategically about my startup, BirdDog.

We’re at the very, very beginning of an experiment with Product Led Growth (PLG).

I’ll explain what that means, how it contrasts with what we’ve been doing (Sales Led Growth), why we’re looking at PLG… and yeah, we’re still bringing up the AI bubble, this time in the context of how PLG further levers our AI Bubble short.

[Curing Neuroticism]

A mentor of mine who built a unicorn always framed running a startup as a problem of progressive growth; if you have character flaws or shortcomings, they'll keep manifesting as problems facing your startup throughout your journey.

Dealing with those problems directly helps both you and your startup grow and thrive.

Now, this DOESN'T imply:

1. Startups are the ONLY want to self actualize (they're not)

2. Building a successful company means you've successfully self actualized (also definitely not; I would suspect that big, founder led companies are dysfunctional in a way that mirrors the founder's own ailments)

All that being said, I think it’s true that a startup tends to get stuck where it’s founders do. And right now, I think BirdDog is stuck on one of my personal flaws - I have a bit of a fear of delegating.

And I don’t mean between myself and Jack; I mean to our customers and our software. I have a habit of doing things for customers when they have an issue, rather than telling them how to do it in the software.

And since I can often times do it faster by writing a sql query or running a python function on the backend than by using our front end, we’re not putting as much of a load on the ui as we could, either.

You’d think this would be a quick fix - just stop doing things you don’t need to do!

In some ways, you should be right.

But it also depends on more narrowly defining what our software does and doesn’t do and getting over a penchant I have for people pleasing. The former is a business level design problem that touches marketing, sales, and product that I believe we basically need to solve to make PLG work; the latter is a personal problem I’m going to solve with the free therapy known as caring more about the outcome of your startup than your own discomfort.

[Knee Capping Asymmetry]

The problem with our current business model is that we’ve artificially throttled our own ability to grow and capture upside.

We’ve been doing Sales Led Growth (SLG). This means that before anyone touches our product, they talk to us on the phone and we make sure it’s a fit and sell them on it. The positives are:

• We can make sure they’re a good fit for the product

• We can charge more and sell for longer terms

• We can and do invest time in training them and configuring the system to help them get the best results

The cons are:

• We create the expectation that we’ll be very hands on with support. This feeds into and enables my neuroticism, because it means I just give them more of the attention we train them to depend on.

• Our net new revenue is limited by the number of sales calls times our win rate times price. Since time input in SLG scales with number of customers, when we get more customers, we have less time. So, we raise prices. When we raise prices, the number of calls we book and deals we win goes down. This leads to less revenue and slower growth.

• We solve more issues with support rather than product improvements

All of this together conspires to kneecap our asymmetry. Meaning, instead of being able to capture an increasing amount of value for every unit of input, we’re capturing a linearly increasing level of value.

On the other hand, a PLG approach stands to let us treat the product as even more leverage than it is now and helps us compound in a bit more of an aggressive way.

[PLG]

To contrast Sales Led Growth, Product Led Growth (PLG) means you let your customer buy your product without even talking to you. This means that product has to be intuitive and easy to use an get value from, like an AI Chat bot or a social media site.

Pros of this approach include:

• Buyers can evaluate the software for low cost and low risk and seriously low marginal effort on our part

• If built right, the product itself will be doing the upselling: meaning, if a user likes it, they will be able to pay more to use ‘more’ of it

• You don’t have to invest time in selling each net new customer

Cons of this approach include:

• Have an entire cheap and nit picky class of buyer that you can otherwise filter out with an expensive sales led process

• Churn will likely be higher

• There will be lots of support tickets, some of which will be real, others which will be total nonsense

• We need to design a system that can scale with spikes in usage (truly, a champagne problem)

These are real trade offs, but we’re making them.

[Building The Mouse Trap]

Now, quite honestly, I don’t know shit about PLG. I’m going to share how I’m thinking about building the solution, but if you have experience doing this sort of thing, and think my framework is wrong, please email / call me.

So far, the framework we’re thinking with is:

• Start with the desired outcome of the users during a session

• Figure out what decisions they have to make to get that outcome

• Design a pretty front end that makes it fun and to make those decisions

• Align incentives so that as they get the valuable outcome, they’re willing and able to pay you for more of that outcome

Outside of that, we’re learning on the fly. I’m lucky & fortunate that Jack has more of a product mind than me, but we’ll still take all of the help we can get, so seriously do write me.

[Grass is Greener]

I don’t think that we’re falling into the grass is greener on the other side problem, but we’re aware we might be.

Outside of our draw to asymmetry, I think PLG better plays to our strengths:

• Jack’s marketing is second to none, and we get more eyes than us two could ever convert with a sales motion

• We have a unit profitable product with differentiated data, so we don’t have to worry about asymmetric loss as we scale.

• We can pretty cleanly tie usage to price and can bake in a lot of opportunities for upselling that are tied to the end user being able to “do their job”

• Our space has a lot of software that has high barriers to entry that doesn’t work; plg derisks this for the buyer

• We want to be positioned to capture market share with considerably lower effort as our “competitors” keep exiting the space (big macro point, see The Bubble Short below)

• A constraint we’re accepting is one on headcount. PLG is much more friendly to a team that doesn’t want to hire than sales led growth is.

All that being said, we’re not jumping head first into the shallow end, either. We’re testing our hypothesis over some weeks.

And it’s not a “pivot,” either. It’s really just making our software easier and more intuitive to use along with experimenting with lowering the barrier to entry.

If you enjoy my sometimes startup sometimes AI sometimes philosophy sometimes jiu jitsu sometimes writing sometimes reading related rants, give it a subscription below - I am here every week.

[The AI Bubble Short]

You might be seeing how this is starting to lean into the ai bubble thesis: I believe that most (if not all) of our ‘competitors’ have designed LLM dependent products that are losing money. 

One such competitor actually had the audacity to ask us how we made our product profitable given inference costs. Now, that’s not something you usually want to go out and answer, certainly not publicly. But I decided that it’s time to spill the beans: BirdDog is profitable because I’m autistic enough to ignore everyone’s advice about how to build a product right now while not being quite so autistic that I can’t make a convincing argument in favor of these decisions and actually execute on it. 

Being a low or negative margin “AI” business right now is actually a much more precarious spot to be in than I think most people realize. In effect, if you’re losing money and your service depends on LLM inference, you’re losing money at an inference price that the provider (ai lab) is losing money at that THEIR provider (data center) is losing money at. 

If everybody in the chain is losing $2 for every $1 being made, the total loss for $1 of revenue in the chain is $8 (2^3). Meaning, the whole supply chain needs outside money to be remotely functional. 

Either the math needs to change DRASTICALLY (which, there’s no indication that it will) or their needs to be infinity dollars. 

Otherwise, when somebody runs out of capital to burn, they have to raise prices. That will more rapidly dry up the already diminishing capital of the other members of the interdependent chain by making them somehow lose more money than they already are. Eventually, somebody else breaks, etc. 

If you’re not convinced by my three paragraphs, please just go read Ed Zitron, whose research is the biggest breath of fresh air in a space that is so far detached from earth they might as well be picnicking on mars. 

Now, for the connection to PLG: we built BirdDog in a way that’s profitable and largely insulated from the bursting of the bubble and price hikes, and we genuinely believe the product is super valuable. 

In this way, the business functions as a short: as “competitors” start going belly up as the bubble bursts, we will be there for their customers to come over to us. 

And, now, we want those customers to have as little friction as possible when they’re ready to check us out.

Hence, BirdDog going PLG is an ai bubble short. 

Live Deeply,