The Symptoms and Root Causes Of Failing Data Efforts

Written By

Crystal Widjaja's (Reforge Partner) story is pretty amazing. She joined Gojek (one of the largest super apps in SE Asia) when it was at 20K orders per day. Over the course of a few years as the SVP Growth and Business Intelligence, she helped it scale to 5M orders per day. To give you a sense of that scale, Gojek completes more daily food orders than Grubhub, Uber Eat, and DoorDash combined plus more trips than Lyft per day 😳. She recently wrote a post on Why Most Analytics Effort Fail with some counterintuitive advice. Here is the quick summary:

Symptoms of Data Issues

It's important to recognize the symptoms of data issues, and separate these from the causes:

  1. Lack of Shared Language - When people describe the same experience in different ways using different terminology. This causes disconnects and time to understand and discuss data.

  2. Slow Transfer of Knowledge - When it takes more and more time to make someone new fully productive. Compensating with more training, is like trying to fix a bad product with more onboarding screens.

  3. Lack of Trust - When people in the org don't trust the data.

  4. Not Being Able To Act on Data Quickly - The longer it takes to get the data, understand it, and discuss it, the longer it takes to act which leads to the data being used even less.

Root Causes of Data Issues

To fix the symptoms, you need to fix the root causes.

  1. Tracking Metrics vs Analyzing Them As The Goal - Many teams view the goal of data initiatives is to track metrics. The real goal though is to analyze those metrics. Those two things are very different. The latter is how we make information actionable.

  2. Developer vs Business User Mindset - A core principle of building any good product is deeply understanding and empathizing with your target user/customer. When building data systems many teams lose sight of who their customer is, or don't have one in mind at all - the business user.

  3. Wrong Level of Abstraction - One of the toughest things to get right with tracking is the right level of abstraction on what to track. Bad tracking is when our events are too broad, good tracking is when our events are too specific, great tracking is when we have balanced the two.

  4. Written vs Visual Communication - Good teams will at least have a shared dictionary that is updated with consistency. But great teams combine visual communication with the written.

  5. Data As A Project vs Initiative - You have to treat your data systems as product that you constantly iterate on. Over time your product will change, your goals will change, and the business changes. As a result, if you aren't constantly iterating it results it results in the Data Wheel of Death.

Crystal goes on to share her step by step process of how she thinks about instrumenting data, along with an event tracking dictionary template here.

Build A Growth Machine

-  Unlock growth for your business, product, or yourself
-  Learn how to never run out of new growth ideas
-  Learn how to rapidly test and analyze experiments
-  Learn the 5 key foundational elements for growth

Headline

Comments