Hello! and welcome to the first Product Update from the Infer team!
In this, we will cover the core flow of the platform and some specific features – giving a taste of the first step in our journey to reimagine exploratory data analysis and research. We will start with a small demo taking you through the flow of the platform before digging in to three feature highlights.
We are still a few months away from launching our public BETA but if you are interested in joining our early private BETA (which is launching this month) please fill our BETA Signup Form. Or forward it, if you know someone else who might be interested.
Previously, in our series "The Beginning of Infer" we covered the motivation and inspiration for Infer and what we ultimately want to achieve. If you want to understand a little bit more about Infer before digging into the specifics of the product.
Let's start by taking a tour of a typical user flow within the platform – from the creation of a project, through to the analysis of the data and, finally, the conclusion of the research.
This tour will cover the core flow and concepts of the platform. There are many more features that we won't touch upon now but dig into in future updates. However, if you are impatient, you can read about some of them in our product documentation here.
The flow of the platform intends to mirror the typical concepts and stages of analytics and research projects: data management, analysis and results.
In our case, data relates to a project, which captures a single overarching piece of research, or analysis. A project brings together a set of analyses joined by a common data set and aim. Practically speaking, each project is made up of a dataset and a set of queries, where the dataset is the subject of each query. Examples of a project aim could be:
The set of queries within a project make up the analysis related to the dataset and the project. Each query should be viewed as either an intermediary step towards a conclusion or a conclusion itself. For example, a conclusion could be a query that identifies the drivers of churn, predicts churn in coming quarters or identifies particular groups of customers at risk of churn and an intermediary step could be one that defines churn, builds new features for analysing churn, learns and predicts churn on the entire data set for other queries to analyse.
By running a query a result set is created. A result set is a table, the output of the query, accompanied by a visualisation specific to the type of SQL-inf command run, eg if predicting the value of a continuous variable, a scatter plot will be shown, or, if predicting the sentiment of a text column, a bar chart of predicted sentiments will be displayed. Each result set can be analysed within the platform and exported either as raw data, for further analysis, or as images, for presentations.
Currently in the Infer platform a project is 1-to-1 with a single uploaded data file. We will in the near future be adding the ability to use multiple datasets and external data warehouses as well. You can upload the same dataset many times and differentiate the projects by giving them different names and descriptions.
In this quick video tour we will showcase the concepts and components mentioned above: creating a project from a data file, creating and running a query, analysing the results, iterating on the analysis and exporting the results.
SQL-inf allows you to easily perform several powerful methods of text analysis on any column of text in your data set. Currently SQL-inf supports sentiment and topic analysis out-of-the-box. Textual analysis can also be chained with other commands to better understand the drivers of the outputs of the text analysis. For example, chaining prediction and explanation with sentiment will allow you to understand the drivers, if any, of a certain sentiment.
Analysis templates give you a quick and easy way to create and run a standard piece of analysis, like the prediction of a column value. The platform currently supports 3 types of templates (Explain, Predict, Similarity) with more to be added soon.
To get your started, the first time you login you will find 3 demo projects already loaded - each with a dataset and a set of queries for you to try out and use as templates, or skeletons, for your own analysis. The three projects cover how to do predictions, explanations, aggregations, groupings and text analysis in the context of churn analysis, customer life time value analysis and analysing unstructured customer feedback.
As mentioned, there are a lot more functions and features in the platform that we will touch upon in future updates. There are also a lot of other new powerful features in the pipeline to be released either as part of the public BETA or for the first public version. As a sneak peek, we wanted to mention three of these in particular.
With the BETA release we will also be releasing a REST API. The API will be free for any user to use. Every Infer account will have the ability to easily generate an API key and use the Infer API to execute SQL-inf commands from their environment of choice. For example, from their data pipelines or frontend analytics tools.
On top of the API, we are also building a native DBT integration for SQL-inf and Infer. This integration will enable users to easily use Infer command within their DBT models and as part of their DBT batches – executing and saving the outputs of SQL-inf commands into their DBT model outputs. All the integration will require is an Infer account, matching API key and the installation of the Infer DBT adaptor package.
We are also bringing in support for BigQuery as a data source. BigQuery will be the the first external data warehouse that we will be supporting. The support will enable you to create projects based on datasets defined on BigQuery tables. This will be followed on by other data warehouse connectors to become our first major premium feature.
And there are many, many more things on the drawing board and in the pipeline. To mention a few: the ability to create forecasting models, more data warehouse connectors(Redshift, Snowflake, etc), more application connectors(Looker, Tableau, MetaBase etc), even better visualisation and analysis functionality, collaboration and user management, better similarity and clustering methods and some incredibly exciting no-code style editor features 🤯
A look at the Infer App - projects and data on the left, analysis at the top and results at