Competing poorly with cat videos and unboxings

Q: Claude.ai, in 280 characters, what will readers get from reading this email that they could not get from spending the same amount of time watching product unboxing videos on Youtube? Be honest. Be honest.

Alrighty then!

A Favor

I'm setting up some new email infrastructure for recruiting research participants so I can do a better job of speaking from the perspective of data to the question of how to intercept more opportunity. This initiative won't turn into benefit for my email list subscribers for another 6 to 12 months, but I could use your help now. Over the coming weeks I'm going to email you all from a new email address ([email protected] or [email protected]). If you'd be so very kind as to reply to that email when you receive it, it'll help establish some positive reputation for the new email addresses. Thanks in advance for any help you are able to lend this.

An Experiment

I just finished a very narrowly-focused competitive analysis for a client. He was curious if any of his competitors have priced and packaged services, or even publish their pricing at all.

This was the perfect kind of research question to bring in a "LLM research assistant" because we were starting with high uncertainty and could tolerate a significant degree of inaccuracy in service of reducing that uncertainty.

Here's what I did:

1. Pulled the names of 1,670 competitor firms from LinkedIn Sales Navigator (LISN) and Apollo

2. Scraped the home page of every firm's web site

3. Wrote some tooling that sent each home page to GPT-4 and asked it to determine if it's a services firm or product business: 720 services firms. Also asked it to determine if it's a generalist firm or specialized solely on my client's vertical: 234 specialized.

4. Scraped every probably-a-services-description page from those 234.

5. Sent those pages to GPT-4 and asked it to look for pricing and evidence of service packaging. Aside from a few who published PPC management fees and a few low-cost productized services, none published pricing or had packaged services for significant scope projects.

It's impressive to me how generative AI makes getting researched answers to questions like this so much easier. A few learnings to pass along:

• One of the fastest ways to size the supply side of a market is via LinkedIn Sales Navigator. Inside of 5 minutes you can have a pretty good ballpark picture of how many companies there are focused on serving a particular market. But some categories will have both product and services companies showing up in even a very tightly defined search. That's why I had to have step 3 in my research process above.

• My "quick, Philip, estimate how many companies in market X are specialized vs. generalist" answer is: 20% of the market, give or take a bit, will be specialized. You'll notice the market I investigated above has 32.5% of its companies specialized (again, according to GPT-4's inspection of the company's home page, so there is some uncertainty baked into this number). That's on the high side of 20%, but the form of specialization I asked GPT-4 to look for was vertical specialization, not a simultaneous vertical and horizontal specialization.

Some Rules

You can approach an email list with an "anything goes" attitude and do sorta OK (ask me how I know hahahahahah cough cough cough...), but it's helpful to have rules you try to live by in your husbandry of an email list, if only to occasionally delight yourself by breaking one of your own rules. That is also an excellent reason to keep these rules to yourself, but I'll make an exception here, just for you, so you'll know how I intend to approach this email list in 2024 and beyond.

• Fewer or no thought experiments/good ideas/theories; more reporting on real-world experiments/research

• The only goal for the list is usefulness at the individual email level; regularity or consistency can't matter or I'll get too wrapped around my own axle and the #1 goal will be threatened

• Occasional promotion of my services, but the marketing for those is simply reporting the past performance of those services, not optimistically selling you on their future potential

• Never ever, ever, ever write “stay tuned for part 2 of this email”

Speaking of thought experiments, the Wikipedia article on Antipodes actually has some interesting examples of what happens when all you have is thought experiments:

Both of these are laugh-out-loud funny from the modern perspective, but when I take a more sympathetic posture, they both strike me as totally reasonable thought experiments that ended up reaching silly conclusions because the experimenters were working with very limited information. Maybe Pomponius Mela thought the Earth's temperature rose exponentially the closer you got to the equator, meaning that you would burn alive if you got too close? Or maybe he thought the equatorial temperature extremely-hot-but-bearable but the journey with limited technology seemed as impossible as a manned journey to the Andromeda galaxy seems to us now.

These were just thought experiments, though, not actual in-the-world experiments with real people, boats, and supplies. The thought experiments only take you so far.

From the most recent "For Fun Poll":

:)

I hope your 2024 is off to a good start.

-P