How Auto Insights Uses Advanced Analytics, AI, and Machine Learning
A lot of people ask us what parts of Auto Insights use advanced analytics, AI, or Machine Learning - this article covers everything about Auto Insights' core capabilities.
How Auto Insights Works
There are 5 main features that make up Auto Insights' core capability:
Pencil–Auto Insights' data contextualization algorithm. This is where data is uploaded, and analyzed and contextualized by Auto Insights.
Search–Auto Insights' ability to process natural language into analytics topics.
Discover–Auto Insights' ability to automatically create insights on what's changed in your data without human input.
What Caused This?–Auto Insights' ability to determine which data points are most significant or important.
Unexpected Changes–Auto Insights' ability to detect anomalies or unexpected events based on historical trends.
These features are built on 4 advanced analytics capabilities. We cover each of these topics in a little more detail below:
What is Data Contextualization?
Raw data is just numbers and often, it doesn’t make sense to people. Humans have a world model that we operate in. If a person sees data with columns named Country and City, they automatically know that both columns are about places and City has a hierarchical relationship with Country. Data in its raw form does not have this ontology and taxonomy.
Auto Insights uses data contextualization to build models over the data, allowing for it to understand the data from various aspects such as:
Data quality–Does the data look clean? By clean: does it contain missing values? If some columns contain high percentages of missing values, Auto Insights will call this out and provide its recommendation (to remove).
Natural language–Data readiness for natural language: Does the data contain natural language materials or is it all codes that are unfriendly to business users? If the data contains a lot of codes, Auto Insights will call this out to the user and suggest either renaming or removing the columns in question.
Relationships–Auto Insights understands correlations and relationships that exist between columns in the dataset.
Like a human, it uses the information in decision-making, later on, to answer questions, guide users to value, and build stories from the data.
What is Natural Language Processing?
The graphic below explains how Auto Insights uses NLP.
What is Pattern Analysis?
There are two areas of Auto Insights that use pattern analysis to provide insights to users - Unexpected Changes and Search.
Unexpected Changes: Pattern analysis is used to detect anomalies or unexpected events from a trend. For datasets that span more than 18 months, Auto Insights is able to perform deseasonalized analysis as well.
Search: When you ask a question such as sales by the department overtime, Auto Insights looks at the trend for each department and is able to call out patterns among them e.g. Do they perform differently, or there are groups of departments that share the same pattern?
What is Guided Journey?
Creating a Guided Journey in Auto Insights allows non-technical users to get relevant insights quickly and explore their data easily. There are two areas of Auto Insights that use advanced analytics to create a guided journey for users:
Discover–Auto Insights uses the raw data uploaded via pencil to automatically generate a report highlighting key changes in the data, what's driving them, and unexpected changes associated with them
What Caused This?–Auto Insights uses models to understand what data points are most important or sensitive, based on what’s in your data. However, it’s important to note that Auto Insights can’t make inferences - the drivers have to be in your data for Auto Insights to be able to analyze them!