Regions

Channel now uses Azure OpenAI to run queries in the EU (France), Australia, Japan, and the US. So, for users of Channel who need to keep their data in a specific region, you can now do that. Also this gives us better performance for users in those regions. Let us know if you have any questions about this or if you want to use Channel in a region that we don't support yet.

Improvements

• We fixed a bug with Datasets where generating column descriptions would sometimes spin forever.
• We stopped allowing duplicate columns in datasets.
• We fixed a bug where images wouldn't load in the EU region.
• Header nodes are now much easier to click on.

We've made a lot of big changes this week: we've introduced Datasets and the Clarification Flow, and we've redesigned the dashboard.

Datasets

A Dataset is basically a big table that you can use for reporting in Channel. After you add a Data Source (e.g. a Snowflake warehouse, a Big Query instance, etc.) data users can create a Dataset by writing a SQL query to define it, adding some description for what it should be used for and then sharing it with the rest of the workspace. You can create as many Datasets as you want.

To create a good Dataset, you should first think about what sort of questions you might want to ask and then join together all the tables you think you might need into one massive table that you can then use to answer those questions. For instance, if you're an eCommerce company and you want to know how many orders you've had in the last week, you might want to create a Dataset that joins together your orders table, your customers table and your products table. A good dataset should have thirty or so columns, and should be able to answer a lot of questions.

The Clarification Flow

Channel is all about asking questions about your data. We've made it easier to ask questions by introducing the Clarification Flow. Instead of picking a Data Source when you ask a question, you now just ask your question and Channel will start asking you questions so that it can really nail down exactly what data you're looking for and where it should look for it. For instance, when I ask something vague like "what is the trend in sales", then Channel will ask me "over what time period, what granularity, and how exactly should I define sales". Once I've answered all of those questions, Channel will then go and find the answer for me.

Through this process, Channel will also choose which Dataset is the right one to use. This means your business users can just ask questions without worrying about the underlying data, and Channel will figure out the rest.

Dashboard redesign

We've also redesigned the dashboard to make it easier to ask questions and tell stories. Most of the tools are now on the top left of the screen, and we've added a new datasets tool that lets you search through your datasets and preview them. The goal here is to make it easier to understand what sort of questions you can ask.

The blue "Ask a question" button is gone now too, and instead you can just start typing in the bar at the bottom to ask questions. Hopefully a lot easier to use!

Improvements

• We made the titles on dashboard cards a little smaller.
• We fixed a bug were SQL "previews" were actually triggering full queries instead of limiting to a fixed number of rows.
• We made the handles on dashboard cards a little smaller.
• We made a few improvements to the SQL editor: better performance, better error messages.

Our SQL editor got some love

Our SQL editor already had AI powered auto-complete and syntax highlighting, but we've improved the performance and accuracy of the auto-complete and made the syntax highlighting a lot more robust on complex Big Query queries. We've also added error detection and highlighting to help you find and fix errors in SQL faster.

Improvements

• We've changed the download CSV experience so that once you trigger a download the CSV will be automatically downloaded once it's been created, rather than having to click another button once it's ready.
• You can now specify your timezone in the workspace settings page (here) and we'll use that to display dates and times in the correct timezone for you, and generate the correct answers for your timezone when you ask questions about dates and times.
• We improved how Channel searches columns for values, by giving Channel the ability to do fuzzy matching on your data.
• We no longer automatically refresh an answers data when the answer is hidden.

Quite a lot has changed since our last changelog. We've been working hard on a lot of new features and improvements, and we're excited to share them with you!

Confidence indicators

Channel will now show on the dashboard which examples and definitions have been used to generate an answer, along with whether or not the answer has been verified by a data user. The goal for us here is to make it easier to understand how Channel is coming up with answers, and to make it easier to double check answers that are important to you. Please let us know what you think!

We have a new SQL Editor!

Writing SQL is a critical workflow for data users and we've been working hard to make it as easy as possible. We've shipped a new SQL editor that has table stakes features like syntax highlighting, dot completion, and a results viewer. We've also added AI powered autocomplete that suggests columns and tables as you type, using the verified examples that you've given us. We're going to be adding a lot more features to this editor soon, so stay tuned.

Improvements

• We've shipped an entirely new Answers page. It breaks down answers into For Review, Confident, Verified and Hidden. The goal is for data users to be able to quickly see what's been double checked and what hasn't, and for business users to be able to quickly see what's been verified and what hasn't.
• We have a whole new coverage page for data sources. It's a lot more detailed and gives you a lot more information about what Channel knows about your data: which columns, tables, schemas it has seen before, some aggregated statistics, and a lot more. We're going to be adding the ability to ask targetted questions directly from this page soon.
• You can leave comments on dashboards now and @ people on your team to notify them.
• We have notifications! At the moment it's just for when you've been @'d in a comment and when you've had something be double checked, but we'll be adding more notifications soon.
• You can now lock dashboards to prevent accidental changes. This is useful if you want to share a dashboard with someone but don't want them to be able to edit it.
• You can copy a link to a particular question or section of a dashboard to communicate with your team more effectively.
• Channel knows what your default currency is, so when you ask questions about money you don't have to specify the currency. If you want to use a different currency, you can specify it in your question and Channel will ask you what the right exchange rate is.
• Question titles are now a lot bigger so you can see them from across the canvas without zooming.
• We now support SAML SSO log in, so if this is something you're interested in, please reach out to us.
• We shipped a new onboarding flow that looks back at the query history of your database and suggests questions you might want to ask. At the moment this is only for Snowflake, Big Query and Redshift, but we'll be adding more databases soon.
• We changed how Channel looks through a database column to find relevant values. Now, Channel's AI can optionally specify a search term which lets it be more precise when formulating WHERE clauses.
• We improved the tooltips on dashboards to make them appear correctly wherever they're triggered.
• We fixed some bugs with axis ticks on charts.
• We fixed some bugs when submitting edited SQL, and in general made that experience much better.
• Some UI bugs got fixed in the UI we use for searching for questions asked by specific authors.
• We fixed some NaN bugs that happened whenever Channel picked the wrong data type for a given column.

Improved definitions workflow

We've heard from a lot of you that using definitions is a critical part of your workflow with Channel, but that it wasn't very intuitive to create definitions, nor to see where Channel was using them. We've completely redesigned the definitions page to make it easier to create and manage definitions as well as understand why Channel selects them when answering a question. You can also see who created a definition and when it was last updated, so that you can update them without worrying about changing someone else's work. We also changed the way Channel determines which definitions are relevant to a given question, which makes Channel more accurate when answering questions.

We'd love to hear your feedback on the new definitions page, so please let us know what you think! This is the first step in a series of improvements we're making to make it easier to understand what Channel is doing and why - stay tuned for more!

Improvements

• We completely redesigned how collaboration on dashboards work in the background which should mean you can have more people in a dashboard at once, build bigger dashboards, and have less issues with dashboards not updating.
• Channel will now auto-name your dashboards after a few questions if you leave it untitled. This is still editable, and is designed to make it easier to find your dashboards later.
• We improved the command palette so that you can now search for previous questions, and Channel will suggest questions directly in the palette too.
• Channel will now fit the y axis of charts to the data, so that you can see the data more clearly.
• We fixed a bug when editing the SQL that Channel generates for an answer.
• We improved the way Channel searches for examples and definitions.
• We fixed a bug where the question you were entering would disappear if you changed data source.
• We fixed a bug where holding a tool shortcut could create multiple copies of the tool, e.g. if you held down N then you would get multiple new notes.
• The UI for questions now doesn't change size depending on the data source.

SQL Explainer

SQL is already pretty complicated, and Channel can generate especially complicated SQL if you try. So, we've added a new SQL explainer feature which will help you understand what Channel is doing under the hood, in plain English. For every answer, you can go to the SQL tab and hover over the "What does this code do?" button to see a plain English explanation of what the SQL is doing. Let us know if you find it useful!

Improvements

• You can now have as many bars as you want on the bar chart and the chart will horizontally scroll if needed.
• We removed the "Test Connection" button from the data source pages. Instead, when you create or update a data source, we will automatically test the connection for you.
• This test now actually works! We test everything we need to test to make sure the connection is valid, and we will show you a useful error if it fails.
• You can now use the command palette on the dashboard (press cmd+k) to search for previously asked questions as well as new ones.
• We fixed a bug which caused tabled not present in the dbt schema to be ignored.
• A bug where Channel would fail to find tables because of case sensitivity has been fixed.
• We fixed some bad links on the answers page.
• A load of visual bugs got fixed on question visualisations. In particular, axis ticks shouldn't overlap the legend anymore.
• The sign out button now works instantly instead of having a little delay.

Channel now supports public dashboards! When you click on the Share button on a dashboard you'll see some new options to make a dashboard public and copy a dashboard with your current position. When you make a dashboard public then anyone with the link will be able to see but not edit the dashboard which is great for sharing with your teammates that don't have Channel accounts or clients that you want to present to. In the next few weeks we'll be adding password protection, the ability to share dashbaords with specific emails, and a more fully featured commenting and notifications system to help you collaborate with the rest of your team.

Interact from the dashboard

This week we also added the ability to interact (i.e. ask follow up questions) from the dashboard. When you click on a visualisation card you'll see another button at the bottom of the card which let's you talk to Channel about that specific visualisation. This is great for asking follow up questions or getting more information about a specific data point, without having to go back to the modal.

Improvements

• We added a new Questions page at /questions which shows all the questions you've asked Channel, which ones are double checked and which ones your teammates have requested double checks for.
• Line charts now correctly calculate the axis limits when there are negative numbers.
• We fixed a bug where axis ticks wouldn't change when you changed whether percentages were premultiplied.
• We fixed another bug where percentages would, by default, be 100x too large.
• We added the ability to un-double check answers.
• You can now double check questions directly rather than having to be on the dashboard.
• We fixed a load of bugs in our onboarding flow.

This week we've added the ability to add the underlying data of a visualisation to the dashboard. You click on a visualisation, then two buttons appear at the bottom of the card and then by clicking the left one you can add a table of the underlying data to your dashboard. By clicking the other button you can see the other relevant questions that Channel generated for you.

This week, we'll be adding the ability to ask follow up questions from the dashboard, as well as shareable links.

Improvements

• When connecting Snowflake data sources, we now strip the snowflakecomputing.com domain from the host name. This should make it easier to connect to Snowflake data sources.
• Channel now by default pivots data correctly, infills missing timeseries data, and handles null values in a more robust way.
• Channel also by default tries to sort data in a useful way.
• On stacked bar charts, the tooltip is no longer reversed.
• Question stack animations now have a little delay before being triggered, so that you don't accidentally trigger them when you're just trying to move your mouse.
• If you get your password wrong when logging in, we now tell you that you got your password wrong.
• We've changed how titles work - you now give a title to an answer rather than a question, so you should see less Untitled placeholders everywhere.

One of the most confusing parts of our product has been what happens to a graph on a the dashbaord when you ask a follow up question - should the card reflect the latest change or the original question? We've been thinking about this for a while and have come up with a solution that we think will make things a lot clearer: question stacks. Now, all of the follow up questions get stacked one by one on top of the original question. This way, you can see the history of the question and pull out any of the previous questions.

Improvements

• User avatars have better default behavior so less flickering.
• Bar charts no longer have flickering axis ticks when you swap between checkpoints.
• Generated SQL is automatically pretty-printed.
• When creating a Big Query data source, we default to having one dataset ID rather than none.
• When you're editing graph titles on the dashboard you can no longer accidentally trigger keyboard shortcuts.
• We fixed a bug where the compiler would get confused by MySQL or MariaDB schemas (these databases don't have schemas!)
• Scorecards have much better font size scaling now, with variable padding (they look a lot nicer!)
• Dashboard websocket performance should be a lot better.
• The way the compiler checks for enum columns has been improved: instead of just checking for the number of distinct entries in a column we also check for the number of GPT tokens.
• Legends on graphs get truncated less often.
• We improved some of our transactional emails.

We spent this week mostly polishing the Dashboard and improving our SQL compiler. The highlights are that the dashboard is now more performant and intuitive, the SQL compiler is more interpretable and robust, and the whole app should be more stable. In particular, the messages section of the modal has been made a lot more expressive and clear.

Improvements

• We fixed some overlap issues in the dashboard so now the item you're selecting is always on top.
• Multiselection states are now properly synced to other users in the dashbord.
• The Channel Compiler will tell you which examples and definitions it's using when it compiles your SQL.
• The Compiler also deals better with predictions and statistical tests now.
• The SQL editor in the question modal is now more robust and saves your state properly.
• The dashboard modal shows which data source was used.
• Negative numbers are formatted properly on question answers.