Release Notes - 23 December 2022

With BBQ plans locked in for the holiday season 🎉🎉 and a lot of shopping to do, we could be forgiven for dropping a small release at the pointy end of the year. Fortunately, for your data analysis needs, that is not how the Relevance AI team works! 🎁🎁

Step beneath the mistletoe 🎄🎄 and close your eyes - as we dive dreamily into another release rundown 🫶🫶

Revamped workflow library💪💪

Lots of editor improvements🚀 ✏️

Introducing cross-tabs!💡💡

We’ve a lot to cover, so let’s get cracking! 👇

Transform your data with a revamped workflow library 💪💪

One of the great things about Relevance AI is the large library of workflows we provide, allowing you to transform your data without code.

For example, we have workflows for tagging your responses, detecting categories with AI or translating them into English. These are powered under the hood by the cutting edge technology in the AI industry, made super easy to use by our platform.

Problem: what workflow should I use?

One of the issues we’ve seen with our workflows is that it’s difficult to discover, understand and share them. Many of our users don’t even know the full extent of power that lies beneath their fingertips.

Solution: find the right workflow every time!

To address this, we’ve resigned our workflow pages to make them way easier to understand.

We’ve started filling them with details on use cases, what happens under the hood, and how to fully take advantage. Furthermore, you can now view and share these workflows without having to create an account!

Upon starting a workflow, you will be led through the sign up process before returning to your workflow. This makes it easier to share workflows with others in your organization, if you think it may be helpful.

Alongside building more workflows, this redesign signifies our new goal of making sure each workflow is packed with lots of content to help you out!

Lots of editor improvements based on your feedback 🚀 ✏️

In our last release, we launched a new category editing experience that helped you easily move responses between categories (such as AI categories, or tags), as well as merge different categories together. Based on your feedback, we’ve further iterated on these features to save you even more time!

A better tag editing experience

Our initial design was useful for responses that have a single attached category (i.e. our AI category feature), however was a bit unintuitive for working with responses that have multiple categories (tags).

To address this, we’ve made a key visual update. You will now see a list of all the categories the response has.

You will notice some are highlighted in blue 💙

These are the tags that will be changed if you drag the response into a tag on the sidebar. You select a tag (causing it to become highlighted) by checking it in the sidebar.

So, for example - in the screenshot above, if I drag it into the category “internet of things”, the tags “mobile phone” and “analysis” will be replaced by “internet of things”, while the rest will remain.

Quick suggestions to make bulk editing faster

The Rules-Based Editor is powerful due to its ultimate flexibility, allowing you to construct filters to find any subset of responses in your dataset. However, we’ve noticed there are some very common actions our users are taking.

The most common being, filtering for all untagged documents in your dataset. So, we’ve added a button that prefills this filter for speed! We will continue to add new quick filters here based on usage!

You’ll notice that your bulk edits also get suggestions!

If you are filtering a tag field, we will make tag suggestions - and the same goes for AI categories.

As stated, we will continue to add new suggestions here based on what we see our users doing most commonly! If you have a suggestion, don’t be afraid to reach out!

Improvements to moving responses

A few other quality of life improvements have been made to the experience of moving responses between categories.

The most notable being, rather than having to drag - there is also a dropdown that appears in the top right now. You can select a category explicitly from there, which may be easier for you than dragging.


Easter egg:

If you do choose to drag, you’ll now notice the sidebar expands in size to help you see your categories more easily.

Cross-tabulation? Cross it off your wish list 💡💡

One of the most common features requested by Relevance AI users is the ability to set up cross-tabs. We’ve worked hard on building what we think is a great cross-tabulation feature, and it’s finally ready for your dashboards!

What is a cross-tab?


It’s a table that highlights the relationship between two or more variables, sometimes also known as a “contingency table”. It’s a very popular way to analyse the relationship between demographics in market research.

Quick guide to cross-tabs

Getting started

  1. Access the 'crosstab' type in any place you might set up a chart.
  2. In Columns and Rows, select the categorical variables you want to compare. Examples of categorical variables include demographic characteristics such as area, age, gender or income bracket.
  3. Your cross-tab will automatically appear and is now ready to be customised!

Tips & tricks!

Cross-tabs tell stories, but not when they’re difficult to read. We suggest the following to make your cross-tab easier to interpret:

  • Display a maximum of two metrics in each cell. Count and Row Percentages are good to start off with.
  • Use a header to provide context to your cross-tab.
  • Enable insights and over/under arrows to make trends easier to visualise.
  • Only enable row / column totals when they support the data.

Customise your cross-tabs

Add headers

Display a header above your cross-tab to provide context. Perfect for survey questions, e.g. What insurance provider do you have a policy with?

Show/hide counts (observed & expected) and % (row & column)

You can display up to four metrics in each cell:

  • Count (default): the frequency of the cell in your dataset.
  • Expected count: the expected frequency of the cell, calculated against your dataset.
  • Column percentage: the proportion of the cell relative to the column’s total count, represented as a percentage. Useful for breaking down a category.
    Example: If you’re looking at the relationship between age and insurance provider, column percentage helps you breakdown how popular each insurance provider is within an age bracket.
  • Row percentage: the proportion of the cell relative to the row’s total count, represented as a percentage. Useful for breaking down a category.
    Example: If you’re looking at the relationship between age and insurance provider, column percentage helps you breakdown the distribution of ages for a provider.

You can re-arrange metrics by dragging the handle.

Show row / column totals

Displays row / column totals for each cell metric.

Highlight insights (over/under indexing is here!)

Enabled by default

See if cells are above or below their expected value, and whether the two variables you selected are related. Each cell is color-coded: blue means above (over-indexing), red means below (under-indexing). An insight displays at the bottom of the cross-tab showing whether a statistically significant relationship exists between the two variables or not.

Selecting over/under arrows displays a corresponding arrow aside each cell value indicating whether it’s above / below the expected.

Show statistical significant relationships - powered by Chi-squared test

Cross-tab insights are based on a Chi-squared test computed at a significance value of 0.05. Selecting this displays the computed statistic, degrees of freedom and p-value below the cross-tab.

Let us know what you want!

This is just the beginning for cross-tabs and we’d love to hear your feedback. Let us know if there’s anything you would like added to make your analysis easier!

We wish a very merry holiday! 🎁🎁

This will be the last release notes from our team for 2022, and we would like to wish you the happiest of holiday periods.

It’s been a truly exciting and rewarding experience for our team, building unstructured data analysis tools for you. We’ve loved all of the feedback we’ve received, the enablement sessions, as well as your patience with any bugs.

It is cliche’d to say, but 2022 was very much the beginning. In 2023 we will be coming back with renewed energy and a lot of excitement to take Relevance AI to the next level. We look forward to seeing you on the platform in the new year!