Why you should stop building infrastructure products
Why data startups should not get into infrastructure but build e2e products; Why you need to ditch all data best practices; Generative AI will boom over the next decade.
This is the Three Data Point Thursday making your business smarter with data & AI.
Want to share anything with me? Hit me up on Twitter @sbalnojan or Linkedin.
Let’s dive in!
Why data startups should not get into infrastructure but build e2e products
Why you need to ditch all data best practices
Generative AI will boom over the next decade
Why data startups should not get into infrastructure
TL;DR: If you found a data startup, prefer end-to-end products over infrastructure. Otherwise, you will be limited in your growth by external factors out of your control.
That’s the size of the e-commerce market (in 1994, that mainly meant via phone).
Amazon isn’t the giant it is today because it grew to dominate an industry,
It is the giant today because the pie it has a big slice of is now literally a hundred times as big!
… and if you read the Everything Store, you know, this was a deliberate choice based on this exact insight.
So what? And so it is with the data space. We’re bullish on the data space because we believe in the continued exponential growth that in 20 years, the amount of data will be 100x.
Fun fact: that’s literally our projection, 20 years, 100x.
So what’s the problem with “infrastructure data businesses”?
By building infrastructure-based companies, you’re setting yourself up to get only 1-10x of that growth, not 100x.
It’s like if Amazon decided to only sell to distributors or only ship articles. It’s not end-to-end.
The solution? Do what Amazon did. They started end-to-end but focused on a small horizontal segment: books.
There are three reasons you should always try to create (slimmer!) end-to-end products instead over deep infrastructure ones:
The snowflake problem: The data space lacks a standard, and there isn’t an apparent one on the horizon. You’ll need to connect any non-e2e product to a gazillion other solutions.
The growth of data will be 100x. But it’s going to be very unequally distributed. You won’t participate even in 10x if you’re not e2e.
The value of data is almost always in turning it into actions. And that’s waaaay down the road if you start just with “data”. Again, this creates dependencies and means you’ll likely miss most of the growth.
The good news? We’ve yet to see a deep infrastructure product that couldn’t be instead turned into an e2e solution with the same or less effort. It’s just hard work!
Why you need to ditch all data “best practices”
TL;DR: We data writers are lying to you all the time. There are no “best practices” in the data space. There’s only one: you must act to learn what’s right for you.
We believe the data space is a truly complex space and will stay so for at least the next decade.
If you keep on looking for best practices, you’ll stay lost.
It’s like standing in a completely dark room. There’s no point in asking your friend on the phone where to go,...
he might think he knows, but you’re the one slamming your nose into the wall.
Action produces information. What you need to do is simple: You need to move, probe, and move some more slowly. But don’t run!
Here are five ways you can ditch “best practices”:
If you’re a company, you cut all 6-month or longer plans. We’re looking at that big fat “data mesh” pilot project. Don’t engage in projects longer than 3 months.
You shorten your requirement lists from 10 to the three essential ones.
You stop looking for a solution that’s 95% perfect and instead take the 80% one you can implement quickly.
If you’re a datapreneur, you build highly adaptable things! Build fast.
Build things to reduce complexity inside a specific set of companies (maybe just your own!).
genAI will boom!
They call generative AI the "next productivity frontier,"
TL;DR: gAI is here to stay, but just at the beginning! Start a gAI company today, and expect to work hard for the next decade.
Most business value might center around four business functions:
customer ops
marketing & sales
software engineering
R&D
gAI will accelerate workforce automation: Between 2030-2060, half of all jobs (of today) could be automated.
Workers will need support learning new skills, but then gAI can increase productivity by a huge factor.
For a detailed view, I suggest you read this detailed piece by McKinsey; it is packed with good stuff (not just fluff!), like individual sector analyses.
So what? There is a clear path forward for gAI. And it requires a decade of work to pay off big time. You should start now to reap the benefits.