Graphext is a tool for conducting complex analytics without writing code. It helps you transform your data, explore it, visualize it and present your findings from a tidy, browser-based workspace where collaboration is encouraged. We built Graphext to make data science more accessible.
"Without data, you're just another person with an opinion."
- W. Edwards Deming
Your Graphext workspace is divided into teams where you can collect and construct datasets and projects.
Datasets are your raw materials and can be created from files that you upload or integrations you make with APIs or databases.
Projects are a space to turn your raw data into useful insights. Among other things, building a project lets you cluster, enrich or make predictions on your data by selecting a combination of transformative steps.
Teams organize datasets and projects into groups, enabling you to work on specific investigations with specific people.
Inside a project, there are typically 6 panels; Graph, Compare, Correlations, Trends, Details and Insights. These are home to your analysis, each providing a different perspective on your data. The Graph visualizes your data in a network. The Compare panel generates a series of charts exploring the similarities and differences between variables in your data. Correlations lets you study relationships between pairs of variables.
Within the Trends panel, you can use 4 types of charts to visualize the evolution of values over a date range. The Details panel displays your data in a table, allowing you to sort or filter it. The Insights panel is a space to capture your discoveries, providing the ability to customize cards with charts, text and statistics as well as. Importantly, you can filter and segment your data by interacting with the charts and lists that represent your variables within your project sidebars.
Your workspace is divided into teams. Each team has its own datasets and projects panel where you can collect resources for your analysis. These are local to your teams, meaning that the resources you have in one team won't be available from the workspace of another team.
Teams are designed for collaborative data science. They encourage you to organize your investigations into groups where you can perform your analysis alongside the people in those groups.
Teams that you are in appear on the thin left sidebar of your Graphext workspace. Team members are given roles allowing them certain permissions in a team. For instance, a Viewer can't edit projects whereas a Member or an Admin can.
Datasets can be constructed from files you upload or integrations you make with databases or remotely hosted datasets. They get stored inside the Datasets panel and belong to the team in which you added the dataset.
Clicking on a dataset lets you inspect it. You can change the variable types automatically recognised by Graphext as well as managing information about your dataset here.
Projects are built using datasets and like datasets, they are local to the team that you build them in. They are constructed from the series of steps that you compile in the project setup wizard. Each step will customize a feature of template code built into Graphext allowing you to transform your data without doing the heavy lifting. Inside a project, you will find 5 panels to conduct your analysis in.
There are 12 types of analysis for you to choose from when building a project, each intended to transform different types of data in different ways. After choosing a type of analysis, you will move on to customize the kind of exploration you want to do.
The panels in your project are windows into your dataset. The Graph, Compare, Details and Trends panels provide spaces to explore your project in different ways, each offering a different visual perspective on the values in your data.
The Graph represents your data as a network, often clustering similar data points into groups. Each row in your dataset is represented as a node on the graph connected to other data points by links. These links are calculated according to similarities between the variables you want to analyze and make it simple to discover communities or patterns in your data.
The series of charts presented in the Compare panel are designed to reveal the variables that best explain the similarities and differences in your data. First, choose a variable to examine. Next, pick some values belonging to that variable. Finally, browse the charts to understand how those values are distributed among the other variables in your data.
Presenting your data in a table, the Details panel allows you to explore every value in your dataset. You can sort and filter using the Details panel creating samples of your data that you can export. Since building a Graphext project transforms your data, new variables are often added during this process. You can export the transformed dataset from the Details panel.
Trends charts reveal the development of values in your data over time. The 4 types of charts available in the Trends panel of your project, each offering a different way to visualize your data across a range of time. You can annotate these charts to provide extra context to significant points or filter the values shown inside of it to inspect the evolution of specific groups.
In trends charts, date variables are plotted on the x-axis. You can change which variables and values are presented in the chart as well as the way they are counted on the y-axis.
Correlation charts are designed to reveal how the values belonging to one variable are associated with the values belonging to another.
To start mapping correlation in your data, choose a variable and inspect the charts that are generated automatically. Correlation charts work using mutual information and expose the density of your dataset at points where values from two variables meet. Larger and brighter circles inside your charts mean stronger correlation.
The sidebars in your project display your variables as histograms or lists. Interacting with them will filter the data that is presented inside your Graph, Trends and Details panels. You can save the data inside filters as new segmentations. Segmentations that you create behave like variables and are a useful way of collecting a customized group of data points that you can then visualize in the Compare or Trends panels.
Once you've made some discoveries, you can present them inside or outside of Graphext. Alongside publishing your project to a public URL, you can export your charts or create a presentation of the insight cards you've created in your project's Insights panel.
The Insights panel is a space to store bitesize chunks of analysis that highlight significant patterns in your data. Insight cards capture a snapshot of a Graph view or a chart and save it. Pressing the play icon at the bottom of an insight provides the ability to jump back into that point of your analysis. You can then add text, variable charts and statistics to insight cards to clarify or contextualize the information presented inside of it. Inside of the Insights panel you can reorder insight cards or present them in a full-screen slideshow.
You can export elements of your project from any panel inside your project, giving you the ability to customize them outside of Graphext and share them. As well as exporting insights, charts and your transformed dataset, you can export the recipe used to create your project. Charts that you export can be customized before export them, allowing you to control the file type, size and theme of your download.
Publishing a project you've been working on means that anyone can visit a public URL to explore your project for themselves. Although, public visitors won't have permission to make changes, save segmentations or capture insights, they can interact with the panels in the same way that you did when you conducted your analysis. When you publish a project you can control the appearance of the project as well as the panels that are available to view. You can also embed published projects on your own site, giving your site visitors the ability to explore Graphext projects without leaving your page.
We know that data isn't always clean and simple.
Have a look through these topics if you can't see what you are looking for.