Sunny

← All weeks

What I’ve been reading (W37, 2023)

Shreyas Doshi on the difference between measuring and evaluating · Things to learn to understand the realities of investing

[MUST READ] How Linear thinks about customer experience

  • Linear thinks supports an extension of the product and another opportunity to build your brand. That’s why they call it “customer experience” (CX) and not “support”. Whatever the product’s feelings and features, your support experience should emulate that.
  • Linear is fast, so the responses and bug fixes should be quick.
  • The app is designed to be simple and straightforward, so customer replies should be the same.
  • The linear app also includes thoughtfulness, care, and a little magic. It’s the little details that make the product feel magical. Similarly, the little details in a customer interaction make it a memorable experience.

Shreyas Doshi on the difference between measuring and evaluating

“You hear in business meetings that if you can’t measure it, you cannot improve it. These kinds of statements are taken as truths. And over time, I realized used in the wrong context, they tend to cause a lot of damage that we don’t even realize because we feel like we are being smart when we say, “This thing must be measured before we can improve it.” The way I realized some of the flaws in this thinking—of “you cannot improve something that you don’t measure”—is looking at my personal life and evaluating it.

I am a parent. And I asked myself, “Am I measuring how I’m doing as a parent, or am I measuring how I’m doing as a spouse?” And while I clearly improved in certain areas of parenting, I had not done any measurement there.

So then I started asking myself, “Well, if I’ve clearly improved in this area of parenting or this area of being a spouse, but I don’t measure it, I don’t track it on a daily basis, I don’t have any metrics associated with it, and yet I can tell that I’m improving in these areas, … how am I doing that?”

The way I know that I’m improving is because I’m evaluating. And that’s where I feel like there’s a difference between the idea of evaluating how you’re doing and measuring how you’re doing. The fundamental observation here is that sometimes evaluating how something is going is enough. And sometimes if you try to measure a thing, you might improve the measurement, but you might not actually be improving the underlying thing.”

Things to learn to understand the realities of investing

  • Compounding
  • Asset allocation
  • Risk
  • Probability and statistics
  • Leverage/debt
  • Opportunity cost
  • Sunk cost
  • Comparative advantage
  • Impact of fees
  • Behavioural bias

The problem in software engineering is managing complexity, not writing code. (cc: )

  • Lots of applications and tools start simple. They do 80% of the job well, maybe 90%. But that isn’t quite enough. Version 1.1 gets a few more features, more creep into version 1.2, and by the time you get to 3.0, an elegant user interface has become a mess. This increase in complexity is one reason that applications become less useable over time.
  • The problem of complexity isn’t limited to user interfaces; that may be the least important aspect of the problem. Anyone who works in programming has seen the source code for some project evolve from something short, sweet, and clean to a seething mass of bits.
  • Some of that evolution is driven by an increasingly complex world that requires attention to secure programming, cloud deployment, and other issues that didn’t exist a few decades ago.
  • Solve important aspects of software (e.g., security) from the get-go. Designing security from the start almost always leads to a simpler result than bolting security on as an afterthought, and the complexity will stay manageable if new features and security grow together. If we’re serious about complexity, the complexity of building secure systems needs to be managed and controlled in step with the rest of the software. Otherwise, it will add more vulnerabilities.
  • Really brilliant programmers write code that finds a way out of the complexity: code that may be a little longer, a little clearer, a little less clever so that someone can understand it later.

State of data engineering

  • Generative AI, like Chat GPT, is becoming more prevalent in data engineering, although it may not be a natural fit for all use cases.
  • Data engineers will likely be tasked with chunking and vectorising back-end systems, and there may be debates about whether to build a solution in-house or use a SAAS.
  • Apache Iceberg (open table format for analytic datasets) by the tabular.io team is gaining more attention, especially with its integration into Amazon services like Athena.
  • Redshift Serverless now only requires 8 RPUs to provision a cluster, making it a more cost-effective option for certain use cases.
  • dbt (Data Built Tool) is gaining popularity as a SQL processing tool for data pipelines, and there is now a dbt glue adapter available (can be used as a docker image)

Others

Substack edition

View the original post

Shreyas Doshi on the difference between measuring and evaluating

 “You hear in business meetings that if you can’t measure it, you cannot improve it. These kinds of statements are taken as truths. And over time, I realized  used in the wrong context, they tend to cause a lot of damage that we don’t even realize because we feel like we are being smart when we say, “This thing must be measured before we can improve it.” The way I realized some of the flaws in this thinking—of “you cannot improve something that you don’t measure”—is  looking at my personal life and evaluating it.

I am a parent. And I asked myself, “Am I measuring how I’m doing as a parent, or am I measuring how I’m doing as a spouse?” And while I clearly improved in certain areas of parenting, I had not done any measurement there.

So then I started asking myself, “Well, if I’ve clearly improved in this area of parenting or this area of being a spouse, but I don’t measure it, I don’t track it on a daily basis, I don’t have any metrics associated with it, and yet I can tell that I’m improving in these areas, … how am I doing that?”

The way I know that I’m improving is because I’m evaluating. And that’s where I feel like there’s a difference between the idea of evaluating how you’re doing and measuring how you’re doing.

The fundamental observation here is that sometimes evaluating how something is going is enough. And sometimes if you try to measure a thing, you might improve the measurement, but you might not actually be improving the underlying thing.”

How Linear thinks about customer experience

  • Linear thinks supports an extension of the product and another opportunity to build your brand. That’s why they call it "customer experience" (CX) and not "support".  Whatever the product's feelings and features, your support experience should emulate that.

  • Linear is fast, so the responses and bug fixes should be quick.

  • The app is designed to be simple and straightforward, so customer replies should be the same.

  • The linear app also includes thoughtfulness, care, and a little magic. It’s the little details that make the product feel magical. Similarly, the little details in a customer interaction make it a memorable experience.

Things to learn to understand the realities of investing

  • Compounding

  • Asset allocation

  • Risk

  • Probability and statistics

  • Leverage/debt

  • Opportunity cost

  • Sunk cost

  • Comparative advantage

  • Impact of fees

  • Behavioural bias

The problem in software engineering is managing complexity, not writing code.

  • Lots of applications and tools start simple. They do 80% of the job well, maybe 90%. But that isn’t quite enough. Version 1.1 gets a few more features, more creep into version 1.2, and by the time you get to 3.0, an elegant user interface has become a mess. This increase in complexity is one reason that applications become less useable over time.

  • The problem of complexity isn’t limited to user interfaces; that may be the least important aspect of the problem. Anyone who works in programming has seen the source code for some project evolve from something short, sweet, and clean to a seething mass of bits.

  • Some of that evolution is driven by an increasingly complex world that requires attention to secure programming, cloud deployment, and other issues that didn’t exist a few decades ago.

  • Solve important aspects of software (e.g, security) from the get-go. Designing security from the start almost always leads to a simpler result than bolting security on as an afterthought, and the complexity will stay manageable if new features and security grow together. If we’re serious about complexity, the complexity of building secure systems needs to be managed and controlled in step with the rest of the software. Otherwise, it will add more vulnerabilities.

  • Really brilliant programmers write code that finds a way out of the complexity: code that may be a little longer, a little clearer, a little less clever so that someone can understand it later.


State of data engineering 

  • Generative AI, like Chat GPT, is becoming more prevalent in data engineering, although it may not be a natural fit for all use cases.

  • Data engineers will likely be tasked with chunking and vectorising back-end systems, and there may be debates about whether to build a solution in-house or use a SAAS.

  • Apache Iceberg (open table format for analytic datasets) by the tabular.io team is gaining more attention, especially with its integration into Amazon services like Athena.

  • Redshift Serverless now only requires 8 RPUs to provision a cluster, making it a more cost-effective option for certain use cases.

  • dbt (Data Built Tool) is gaining popularity as a SQL processing tool for data pipelines, and there is now a dbt glue adapter available (can be used as a docker image)

Others

Originally on Substack