18
Analyzing Analysts Dustin McIntosh Mode Analytics Understanding the Queriers Behind the Queries

Insight Demo

Embed Size (px)

Citation preview

Analyzing Analysts

Dustin McIntosh

Mode Analytics

Understanding the Queriers Behind the Queries

Goal: Understand user base and how they interact with the product

NovicesPower Users

● Who are the users?

● Which SQL errors do they make?

● How can we make users' experience better?

● Why do users churn?

● Can we classify users' ability level based on their queries?

Goal: Understand user base and how they interact with the product

● Who are the users?

● Which SQL errors do they make?

● How can we make users' experience better?

● Why do users churn?

● Can we classify users' ability level based on their queries?

NovicesPower Users

Time

Churn – Are users getting frustrated?

Churn – Are users getting frustrated?

Churn – Are users getting frustrated?

No!

Churning Users = Tutorial Users

Recommendation: prompt tutorial users to connect their own data at tutorial conclusion

Classifying Users Based on the Content of their Queries

Bag of SQL

keywords

Formatting based:

– White space

– Parentheses

Keyword diversity

Is it an Error?

Error type

Random Forest

Classifying Users Based on the Content of their Queries

Bag of SQL

keywords

Formatting based:

– White space

– Parentheses

Keyword diversity

Is it an Error?

Error type

Random Forest

AUC = 0.66

The differences between experts and novices

Deliverable: model that classifies users' ability based on queries

Dustin McIntosh

EFavDB.com

Extra Slides

What errors do they make?

Recommendation: prominently display or auto-correct table/column names

Top-heavy nature of SQL

Non-tutorial users also not getting frustrated

All churned users Churned users that have connected their own data

Correlations of the Features

Compare Using top 5 features

All features

Top 5 features

Histograms of important variables

novicesexperts