Dr. Mario's 2nd 🧠

We Need a New…database

3 min read

rw-book-cover

Metadata

Highlights

nothing ever quite works this way in practice.4 Tables are rarely that well organized, there are often thousands of them, and they often overlap in confusing and contradictory ways. They’re frequently broken and out of date. And the questions people ask usually return messy answers (View Highlight)

New highlights added 2025-05-18

In other words, most databases are very good, and most data teams are still a disappointment. (View Highlight)

The data of a mid-sized B2B SaaS product simply doesn’t have the potential energy of Google’s search histories, or of Amazon’s browsing logs. (View Highlight)

we assume that there are diamonds buried in our rough data, if only we clean it properly, analyze it effectively, or stuff it through the right YC startup’s new tool. But what if there aren’t? Or what if they’re buried so deep that they’re impractical to extract? What if some data, no matter how clean and complete it is, just isn’t valuable to the business that owns it? (View Highlight)

To be clear, as a resource for mechanical reporting, data is perfectly fine. But it’s hard to go beyond that, because using data is typically an indirect and imprecise way to figure out what you really want to know:

CEOs want to know what their customers are thinking. Behavioral data isn’t “truth;” it’s an observable proxy, the input to a kind of analytical alchemy that attempts to turn individual outcomes into generalizations about intentions. … Though most business decisions are driven by numbers, those numbers matter because they define people’s loose mental models for how the world works, not because people need to know about the often-meaningless tedium of things like statistical significance. (View Highlight)

if someone wants to answer questions like “how can our team move faster?” or “what sales deals are in trouble?,” they have navigate through some roundabout prerequisites:

  1. What quantitative measures would indicate that a sales deal is in trouble?
  2. Using the data that they have, how can we compute those quantitative measures?
  3. How do we interpret our results, especially when most of them are an inconclusively wiggly line? (View Highlight)

comparing the value of 750 customer interviews with a database full of usage data:

Though the [Dropbox folder of customer interviews] is probably more valuable than the [database of event logs], we can’t easily mine or manipulate it; we can only sample it. That’s why we instinctively dismiss this sort of information as untrustworthy or biased: Not because it’s wrong, but because there’s no way to look at all of it at once. But now, with Granola 2.0, you can, quite literally, do exactly that: Record interviews, Granola transcribes them, and gives you a chatbot to query them. It’s not a text-to-SQL-to-proxy-metric-to-an-inscrutable-wiggling-line; it’s just text-to-answer. (View Highlight)

Graph View