Search-based BI is the natural language query capability in Arcadia Data. This capability lets anyone get answers from their data using the familiar paradigm of internet search. The goal of this article is to walk through the setup of search-based BI in Arcadia Instant, our free desktop BI tool available for unlimited use (read on to learn how to get Arcadia Instant). We’ll use one of the included datasets in Arcadia Instant.
Starting with Arcadia Instant
Once you have Arcadia Instant displayed in your browser, click on the VISUALS tab in the menu bar. This gives you a list of all pre-built dashboards and visuals.
Click on the US State Population Trends tile to see an example dashboard using the US State Populations Over Time dataset.
Explore the dashboard by hovering over various data points in the visuals. Note that if you click on any data point, a dialog box appears with the option Filter All.
If you click on Filter All, all visuals will be updated to drill into the specific data point you clicked on. To undo the drill down, click on the Reset button in the upper right corner of the dashboard.
Creating a New Visual using Search-Based BI
Now let’s create a new visual for this dashboard. Click on the EDIT button in the upper left corner. Then click on the Visuals icon in the menu on the right side. Then click on the NEW VISUAL button.
Click on the Search icon in the new visual that appeared. Type the query, population trend by year and state. Note that as you type population, suggestions appear (more on this later). When you hit Enter, see how trend lines appear in the visual.
Notice how some words in your query are underlined. Hover over each term to see how the system interpreted those words. For example, hovering over population shows that it was an aggregation of a table field called population.
This example shows how you can use search-based BI for quickly building visuals, and this is only one way to use search-based BI in Arcadia Data. More uses will be described later, but for now, let’s see how a dataset can be set up for search.
Setting Up a Dataset for Search
Now let’s walk through how the dataset was set up for search. Click on the DATA tab in the top-level menu bar. Then click on the dataset, US State Populations Over Time. Click on Search Modeling in the left panel to get the search modeling page.
Note the checkbox for Enable search on this dataset near the top. This tells the system to use search-based BI on this dataset. When you create a new dataset, this is disabled by default.
Now let’s go over some of the key features in the search modeling page.
In the Fields section, you have a list of database table fields. In the Matching search terms column of the matrix, you can add synonyms for any of the table fields. Note that in this example, the terms time and yr are synonyms for year, so in our previous query example, we could’ve typed population trend by time and state and we would’ve gotten the same result.
Also note the Searchable column. Checking the accompanying checkbox for any table field enables that field for filtering. Since the state field is checked as searchable, all values in that field will be recognized as part of your query. This means you can run queries like population in CA and the value CA will be recognized as a filtering value because it was found as a value in the state field.
The Suggestion Questions section should be self-evident – these are the suggested questions you saw earlier that appear when you start typing in the search bar. This uses a typeahead feature that will match the list of suggested questions to what you start to type.
The Word Substitutions section is useful for mapping abstract concepts into real data points. For example, you might have no concept of “west coast states” in your datasets, but you can define that here. Simply assign west coast states to in CA, “OR”, WA and you can now search on west coast states. Note that I put double quotes around OR so that the system knows it’s a value (abbreviation for Oregon), and not the conjunction “or.” Be sure to click on the SAVE button to save this change.
Now let’s go back to a dashboard to use this word substitution. Click on the NEW DASHBOARD button in the upper right corner. Click on the search icon in the lone visual, and type west coast state population trend by yr and state. You can see that I can use yr instead of year because of the synonym, and I get only states in the west coast as defined by my word substitution.
Search Bar in Dashboards
What we’ve seen so far is how to use search-based BI as a means for quickly building visuals. This works well for a data/business analyst, but what about using search-based BI for non-technical business users? You will typically want to deploy comprehensive analytical applications to end users, but within that application, you can also include ad hoc querying via a search bar.
To do this, click on the NEW VISUAL button. Then click on the pencil icon in the upper right corner of the visual to edit the visual. In the Visual Builder screen, you’ll see a list of visuals in the upper right.
Choose the Search visual, and a search bar will appear in place of the data table. Click on SAVE and CLOSE in the upper left, and now you have a search bar in your dashboard.
Now you can continue building a dashboard that includes ad hoc querying that you can deploy to business users.
Ad Hoc Queries
You don’t need a dashboard to run ad hoc queries per the example above. Just click on the SEARCH button in the upper right corner of the Arcadia Instant interface and you get a search bar that’s ready to go.
For this example, let’s just type state populations. As you type, you’ll see suggested questions as well as your search history appear.
For this query, you can see that two answers appear. The first appears to answer our immediate question, but for an internet search experience, other results will appear as part of the Ranked Answers. If there were more datasets that matched our query, we’d have more potential answers in the results list. You can bookmark the alternative answers to tell the system that they are also potentially valid answers for future similar questions. This will help the system to learn about the datasets, and help all users with data exploration across all datasets.
That’s a quick overview of search-based BI in Arcadia Data. There’s certainly more to explore, so hopefully you can do so while using Arcadia Instant for your desktop analytic needs. If you want to investigate how Arcadia Data can help you and your team at a larger scale, be sure to contact us.