Here’s a great post by Karthik Shashidhar on why they shut down Babbage Insight, and the learnings from the experience.

(I’m reproducing in full here since LinkedIn is hostile to content.)

I added ⭐ to points I found most interesting.

As the more perceptive of you would have figured out by now, we are shutting Babbage Insight . When I told this to one of my old friends, his immediate reaction was “so what were your learnings from this experience?”. And so I decided to write this.

Remember that I’ve written all of this primarily for my own benefit and future reference. Any cost / benefit from this to anyone else is only incidental.

This is also in no particular order. And the salt you need to add is that everything is purely with the benefit of full hindsight.

  • With full benefit of hindsight, this (AI for data analytics / insights) has been a rather tough sector to operate in. Nobody has really cracked it here (not even established players - distribution doesn’t imply traction), and there has been a bloodbath of startups, which (unfortunately) we are also part of
  • ⭐ When you require access to a customer’s data warehouse, it doesn’t matter what size the customer is - every sale is an enterprise sale.
  • From our customers’ perspective (which we belatedly realised), this was a risky buy. They didn’t know what they wanted, and they couldn’t see how what we were offering would clearly help them.
  • ⭐ Hiring is massively underrated, and a massively difficult job. We hired poorly. Maybe the biggest skill required for a startup founder is hiring.
  • Tenure match with your first set of investors. Raise early if you are raising. That way you don’t run out of energy before the money runs out (happened to me)
  • I got into the “where’s my idli” (or “where do i put mine” - both translate to the same phrase in Kannada) syndrome. Every time I read about something someone else in the sector was doing, I got defensive, wondering if someone else is already doing what we wanted to do; rather than figuring out how to learn from or copy the idea. Perversely, despite running a startup, I got into a zero sum mindset (with all my competitors), which didn’t help
  • I should have started with consulting, and then used learnings from that to build products.
  • Even if you are building a product, build it with a specific customer in mind. Building with a generic / hypothetical customer can mean that you build for no one
  • Our attempts to build a “generic enough” product were to no avail, as each customer wanted their own set of customisations. And we made the mistake of trying to tweak the entire product to fit these customisations, rather than hack.
  • We hear romantic stories of people who cracked big-figure cross border sales, but those are driven by spectacularness bias. You need to be in the market that you are selling to. There is no other way. If you cannot, you need to employ someone in that market.
  • I listened to too many people. I tried doing too many things that don’t come intuitively to me, and so struggled.
  • Every single one of my burnouts has come about when I’ve tried to write production grade code. I like coding but software engineering simply demands too much attention to detail for me to do well. And I keep making the same mistake - step in to write code when the team cannot. And burn myself out
  • I love talking to (potential) customers, but I don’t have the discipline for sales. Following up on time. Following through on intros. Etc.
  • ⭐ Hire a small number of highly skilled, even if expensive, people. At early stage startups, especially in the age of AI, people who just implement what you tell them to do are overrated. LLMs imply that you don’t have 10X engineers any more - you have 100X (apologies for that cheesy line). And you need to hire a few of these, else you’ll be stuck.
  • I absolutely hated putting my location as “United States” on LinkedIn, but had to do it for sales purposes. Did a lot of other such things, because they are “done things”.
  • Poor hiring meant I spent too much time writing code, and too little time building the business.
  • I never got a coherent communication of what exactly we are building. Obviously this was a huge problem
  • Our product was doomed the moment I wrote the first line of code.
  • We didn’t use LLMs enough. Our code was too deterministic. This was a function of our hiring.
  • We didn’t vibe code enough.
  • We didn’t move quickly enough. Again it was possibly a function of our hiring.

I’m scheduling this post to be sent at a time when I’m in the middle of my Vipassana. I’m pretty sure I’ll return and double the length of this list!