We’re proud to announce a major product update that we’re calling ... Graphext Fluid.
Data science should be an interactive cycle, giving analysts the autonomy to make reactive decisions and adapt their projects on the fly. To reflect this, we’re making Graphext more dynamic and more flexible so analysts can move faster from raw data to insights.
Automatic Projects - Instantly explore your data with projects created as you upload data.
New Data Table - A smart tabular presentation offering value summaries and ways to grow analysis.
Cast Variable Types - Change the types of your variables on the fly.
New Variable Manager - Group, describe, move, pin and hide columns using variable cards.
Visualize with Plot - Plug your data into a range of common chart types and find patterns (guidance).
Create Graphs & Model Data - Layer on advanced data analysis steps like predictive models and clustering.
"Data analytics is iterative - you check the data, explore it, create models and you move back and forth. Fluid is about adapting Graphext so that it helps analysts to analyze data in a more natural way. At the same time, we’re adding major features that make Graphext more powerful."
- Victoriano Izquierdo, Graphext CEO & Co-founder
Alongside the release of Graphext Fluid, we’re excited to launch affordable pricing options on our website in the very near future. Soon, new customers will be able to pay using credit or debit cards through our website. We’ve designed our new pricing tiers to help you easily scale your plan as and when you need it. But if your just looking to try Graphext out ... our free accounts remain the same and will always be there.
As soon as you upload a dataset to Graphext, we’ll create a project for you to explore its contents. Exploratory data analysis is critical to getting familiar with datasets and involves inspection, visualisation and simple transformation. By creating an initial project for you to work with, we’ve pushed all of the decisions further down the line so you immediately have a chance to get to know your data.
Automatic projects are created with any new dataset you upload or integrate with Graphext. Whether you upload a simple CSV file or set up a PostgreSQL database integration, you’ll get an automatic project to explore. Open up automatic projects and get instant feedback on your datasets using Plot or Compare to visualize simple variable or value patterns.
Tabular representations can be so much more than endless rows of data points. Our new Data table presents value distributions, organizes variable names and descriptions and lets you filter, hide, pin, sort and search for variables.
But more than this, use the Data panel as a workspace to grow and adapt your projects. Data is also home to Graphext’s data enrichments, data transformations and data modelling/network algorithms. In Graphext Fluid, you can add all of these analysis steps to your project on the fly.
You can now change the type of variables inside of your project. Previously, variable types remained fixed inside projects. But as knowledge of a dataset increases, hypotheses can change. Changing the types of variables from inside of a project gives analysts more power to adapt their dataset at any point - a flexibility that can be very useful when applying Graphext’s predictive models to your datasets.
To cast variables, use the Wizard to Start a Project from inside your Data panel. Then, click on the icon next to any variable name and change its type to another from the dropdown list.
We’ve completely redesigned the management of variables in Graphext making it more intuitive to group, describe and organize the columns in your dataset. Our new variable manager brings together everything you need to manage columns in your data into one central hub, featuring draggable interactive cards to add descriptive information and tags to your columns.
Use the variable manager icon at the top of your right variable sidebar to bring up the variable manager, then start organizing the variables in your data. Add tags to group similar variables together throughout your project, hide variables or use the description text box to clarify more complex ones.
As you might already know ... Plot is a space to visualize simple patterns in your data. It contains a range of common data science chart types like histograms, bar charts, time-series charts, box plots and heat maps that you can use to instantly visualize value distributions.
Plot is a fundamental component of Graphext Fluid because exploring data visually is one of the best ways to expose insights that will guide and direct further analysis. Open up Plot in your automatically created project to start visualising key variable and value patterns.
You can now add advanced analytics steps like predictive models and Graphs that cluster your data to Graphext projects - on the go. Rather than building new projects as you level up your analysis, Graphext Fluid lets you add steps like clustering and NLP transformations using the Data panel inside your existing project. This way, analysts can get to know their datasets better before deciding which models to use and how to apply them.
To add a Graph and/or Model to your project head to the associated panel in the top menu bar of your project and click Create Graph or Create Model. This will bring up the setup wizard where you can choose which kind of analysis you want to conduct. Follow the wizard steps to configure your Graph or Model.
Checking the box to overwrite your project will add the Graph or Model to your existing project, whereas leaving the box unchecked will tell Graphext to make another version of your project containing the Graph or Model.
Inside Plot - our new analysis panel - we’ve added more charts to visualise the relationships between one or two variables. With Plot, you can create bar charts, heat maps, box plots and all of the time-series visualisations previously found inside Trends!
Yep ... Graphext’s exploratory features just got a lot more powerful. Here’s an overview of what we’ve added but check the full Plot documentation to learn more.
"There is magic in graphs. The profile of a curve reveals in a flash a whole situation - the life history of an epidemic, a panic, or an era of prosperity."
- Henry D. Hubbard
Plot charts are designed to help you quickly measure value distribution between one or two variables in your data. The different types of charts you will find require different data types - for instance Overview, Compared Segments and Segmented Overview charts require date values.
You can use your sidebar filters to restrict the data presented inside charts in Plot as well as changing the way that data is aggregated or summarized using the dropdowns at the top left of your chart.
Bar charts are a simple representation of two variables. The variable represented on the Y axis must be quantitative. The variable on the X axis can either be quantitative or categorical.
Best used for ... Understanding how values from one or two variables are distributed.
Box plots are great for showing the value distribution belonging to a quantitative variable over a number of different categorical data segments. They represent quartile ranges and median values associated with these categories.
Best used for ... Understanding frequency distribution patterns.
Heat Maps are great at spotting correlation patterns between pairs of variables. They use a color spectrum to represent density of your dataset at points where values from two variables meet.
Best used for ... Measuring correlation between pairs of variables.
Read more about Plot in our documentation here.
This month we’ve built more Trends charts, a new Graph manager and a data enrichment to upsample survey data on top of a number of important improvements.
We’ve added a new type of chart to Trends! Charts showing recurrent patterns are designed to reveal similarities in the way that quantitative data evolves over specific time periods.
These charts look a little like box plot charts and show the median and quartile range of values aggregated by hour, weekday, week, month, quarter or year. Recurrent patterns are great for spotting repeating trends like seasonal dips in property price, stock price increases in January or decreases in email open rates on Friday afternoons.
How can I start using it?
The Graph Manager we’ve just built lives in your right sidebar, making it much easier to see changes you make in real-time. Just bring up the Graph Manager sidebar and watch your actions have immediate effect.
We feel this change makes it much faster and easier to adjust labels or change the size and color of nodes in a Graph.
How can I start using it?
Survey data often needs to be weighed in order to adjust the influence of certain individuals in the final survey estimates. We’ve built an enrichment to scale survey datasets using a column containing predefined weights.
Our enrichment uses the weights already in your survey data to produce a transformed dataset that adjusts the importance of respondents. Read more about the theory behind surveys and why we built this enrichment in our post here.
How can I start using it?
Lists are a special kind of data because they hold multiple values. We’ve added a new metric - Total Count - to tooltips associated with lists. Hover over a list value to see it.
Total Count shows the total number of times a value appears. Everything and Selection metrics refer to the number of rows featuring this value and don’t account for instances where a value occurs twice in one row.
How can I start using it?
- Improved Graphext’s ability to recluster segmentations in big data projects.
- Improved the presentation and organisation of variables in our Text / Social Media > Topics analysis type.
- Add a new variable to Text projects. Length specifies the number of characters in a text value.
- Improved our method of ordering variable charts in Correlations.
- Improved our method of presenting charts without using decimals on either axis.
- Fixed presentational issues with the wizard’s display of creating project steps.
- Fixed a bug causing unexpected behaviour when a user tried to save a manual segmentation.
We’ve been curating big list of the best data media around. Here’s one on our favorite data podcasts. Read more.
A picture of a population is what most surveys hope to achieve. We're taking a look at the fundamentals of survey theory - sampling & weighting - through the lens of a Pew Research survey that examines American attitudes towards relationships and dating apps in 2021. Read more.
This month we've added search bars in our data tables as well as making it easier to explore Compare & Correlations with large datasets. We've also made it easier for you to monitor the progress of your project setup with a new distinction between processing & configuration steps.
Finding variables in large datasets can be frustrating. We've added a search bar to data tables across Graphext to help you find variables quickly and simply.
Start typing the name of a variable inside the search bar in either Details or the Datasets panel of your team workspace. Graphext will automatically scroll to the location of your variable in the data table.
How can I start using it?
We've changed the way that charts are loaded in Compare & Correlations. To make it easier to work with datasets with lots and lots of variables, we've added dynamic loading to these panels.
You probably won't notice much difference here on the face of it but behind the scenes ... this feature means that you can load charts a lot faster even when you are working with surveys that have thousands of questions to compare or correlate.
How can I start using it?
The steps you see as your project is being created are now separated into processing and configuration steps. We've made this change to make it easier for you to monitor the progress of your project setup.
Processing steps relate to the data science steps involved in transforming your dataset and creating your network. Configuration steps relate to the presentational aspects of your project setup. Start building a project and check the new dropdown menu inside of the project execution sidebar.
How can I start using it?
- Fixed a bug making it difficult to filter data using our interactive histograms
- Fixed a bug preventing a user from editing a manual segmentation with expected behaviour.
We spoke to the data science team at Aquaservice about how they used Graphext to build a clustering model to improve the way they forecast consumer demand. Their project grouped delivery routes using over 30 factors to calculate similarity and exposed patterns in the errors made by their prediction models. Read more.
Why are the social media strategies of Innocent Drinks considered as the gold standard for marketing teams the world over? We collected every tweet (10,521) posted by the communication department to deconstruct Innocent's content, style, reach and engagement with a simple topic analysis. Read more.
We've made significant improvements to Correlations! With the addition of relative mode, Correlation charts can be simpler & easier to read because they show the percentage distribution of values belonging to a variable.
Across Graphext, relative mode presents data as a proportional representation. In practise, this means you see data as a percentage distribution rather than an absolute count.
This is especially useful in Correlations charts because these use size and color to visualize the correlation between lots of values belonging to pairs of variables. With relative mode, the size and color range of bubbles in Correlations charts are restricted to a percentage distribution (either on the x or y axis). This makes it easier to spot patterns.
How can I start using it?
Sentiment analysis in Graphext just became much more powerful with our new enrichment - integrated with an industry-leading model from Cardiff NLP & hosted by Hugging Face.
We've also added detailed documentation about our analysis types on our website and in the app!
Our new sentiment analysis enrichment is built using an industry-leading model from the team at Cardiff NLP and hosted by Hugging Face. Sentiment analysis models predict whether text is positive, negative or neutral. Check out the documentation describing the mechanics of the model here.
Choose this enrichment using the Data Enrichment tab in your project setup wizard to start classifying the sentiment of news headlines, song lyrics, tweets and more text of all shapes and sizes.
How can I start using it?
- Open up a dataset that contains a text variable.
- Choose any type of analysis to perform.
- Inside your Data Enrichment tab, choose Analyze Text Sentiment - Cardiff NLP.
- Select the text variable you want to analyze the sentiment of.
- Complete your project setup.
- Once Graphext has built your project, open it up and explore your new Category of Sentiment (Cardiff NLP) variable.
We've started adding documentation to help you make the most of our analysis types. You can find these in the app using the information icon inside the card for each analysis type or in the docs on our website.
We've written this to help you understand the best way to approach each of our analysis types. Expect walkthroughs, use case examples and exact directions on getting started.
How can I start using it?
- Choose a dataset to work with.
- Click on one of the information icons inside of your project setup wizard.
- Click the link at the bottom of the documentation to read more.
- Fixed a bug causing Graphext to freeze after a user saves a manual segmentation.
- Fixed a bug stopping Trends | Segmented Overview charts from presenting text variables.
- Fixed a design issue making it impossible to download datasets from Details when the dataset has an especially long name.
- Fixed an issue stopping relative mode from being available in published projects.
- Fixed a problem with the transfer of information between Insights and Compare
- Fixed a design issue with the alignment of chart legends inside Insights.
Our latest Data Academy instalment looks at how NLP can be used by businesses to analyze text. Starting from text analysis fundamentals and moving on to look at more complex recent developments in the field of NLP, this article is intended to introduce and equip business and data analysts with knowledge, techniques and tools to take forward in their text analysis projects. Read more.
What's the most important milestone in a relationship? According to data from a Stanford study, it's a day like any other that occurs somewhere between the 4th and 5th anniversary of a relationship. We built a simple project with the intention of finding the moment in a relationship where breaking up is less likely than staying together. Read more.
You can now customize the values shown in Compare & Correlations charts. We've added a search bar to help you add important categories to these visualisations.
We've also added a new color palette to your projects that uses a dynamic scale that updates depending on the number of values belonging to a variable.
Up until now, charts in Compare and Correlations presented only the most frequently occurring values from a variable. You can now choose which values to present in these charts using the search bar at the top right of each chart card.
Open up Compare or Correlations and choose a chart with hidden values. Click the search bar from the top of the chart and add in your new value.
We've added a new color palette to your projects. Re is slightly different to Horus and Osiris in that it offers a dynamic scale of colors that will update depending on the number of values belonging to a variable.
Re is particularly useful when exploring a small to medium range of categorical values. Its colors move from light blue through orange and red to purple on a scale that is calculated according to the number of values in a category.
- When you upload a new dataset - Graphext will now be able to tell the difference between Categorical values vs Text values with greater precision.
- Your color palette will now be saved to your project settings. This means that closing then returning to the project will not affect your choice of color palette.
- We've improved the way URL variables are presented throughout your datasets and projects. URL variables will now be presented in the same way that categorical variables are presented.
For our first Data Academy release - we've gone back to basics with Exploratory Data Analysis. This article covers what cleaning, transforming and enriching data means as well as explaining why different visualisation types can be useful for studying different types of variable relationships. Read more.
You can now copy datasets and projects between Graphext teams. We've also made it easier to inspect text or quantitative values in greater detail with new text tooltips in data tables and the ability to save variables that capture zoom-ins on quantitative ranges.
We're also pretty excited to announce that you can now customize the thumbnails inside your project card - using uploads or new Graph captures.
You can now move or copy datasets across workspaces as well as making copies of key projects. Click the menu icon from your Graphext team workspace and choose 'Move to' or 'Make a copy in' to give other teams access to your data and analysis.
We've added this feature to make it easier for you to collaborate on and share important analyses that you create. Making changes in a copied project won't affect the state of your original project.
You now can upload, regenerate or enlarge your project thumbnail images! Head to your project settings, click on the project image and choose how to set your new one!
The size of project thumbnails is set to optimal dimensions - meaning that any image you set is guaranteed to look snazzy!
Zooming in on specific value ranges isn't a new Graphext feature. But up until this point - any zoom-ins you make on quantitative variables will disappear as soon as you reload a project. Now ... they won't!
Zooming in on quantitative ranges helps you account for extreme values in your data. Zoom in on specific ranges to explore data distribution between two points.
We've added tooltips to the table in your Details panel - helping you inspect the full content of text in your data. Hover over a text value to reveal its full content.
You can also copy the content of a text value by right-clicking on it and selecting Copy!
Now you can remove any variable from your project. Click the right menu next to the variable card in your project sidebar and choose Remove from the menu list.
Cleaning up your analysis is a useful habit to get into. Removing a variable from a project will delete any reference to it in all of your project panels.
When filtering data in your projects, your sidebar charts will now jump to Relative mode by default. Relative mode means that data in your selection is shown in proportion to the distribution of values in your whole dataset.
- Added the ability to view and edit the project recipe from the project settings window.
- Removed automatic filtering on datasets of any size so that - by default - projects will be built using the full dataset.
- Fixed a bug stopping labels from appearing when users hover above nodes in the Graph.
- Fixed a bug stopping users from sending data to the trash from panels outside of the Graph.
- Fixed a bug causing mixed JSON data to crash on upload.
- Corrected a problem causing tagged variables to appear in the wrong variable collections inside Cluster projects.
We used sentiment analysis to model 5100 Billboard chart-toppers between 1964 and 2015. Our analysis predicted whether song lyrics were positive, negative or neutral as well as detecting the topic and intent behind the most popular tunes in music history. Read more.