Engine Works
Under the hood of Alteryx: tips, tricks and how-tos.
By Megan Dibble and David Kim (@dimkavid)
Alteryx Auto Insights can add value to your business by uncovering root cause analysis, automating data exploration, and streamlining the reporting process. To realize that value, it’s good to know the best practices for setting up and using the product. Let’s go through some guidance from our Customer Success team based on their deep knowledge of the software!
Problem Formulation & Setup
Before you start loading data into Auto Insights, it’s worth spending some time thinking about the problems you want to solve and the questions you would like to answer.
Auto Insights can answer questions about a dataset like:
- What’s happening?
- What’s important?
- What changed?
- What’s going wrong?
- What caused this?
- What relationships exist?
To answer these questions, you need data with rich categorical dimensions to explore. You also need to identify key measures (KPIs) that you would like to analyze. While Auto Insights can certainly uncover findings in almost any dataset or schema, we recommend a structured and timestamped transactional dataset to leverage the strengths of Auto Insights and identify changes and patterns over time.
To dive deeper into what makes a good Auto Insights use case and see some examples, check out this Community article.
Data Selection and Formatting
Your data might not be in the right format for Auto Insights—yet. Below are some best practices to consider while prepping your dataset.
Use granular data
It’s important to understand whether your data is at a summary or a granular level. The more aggregated each data record is, the more aggregated the insights will be. Depending on your needs, consider uploading more granular data and allowing Auto Insights to aggregate – giving you more flexibility to look at different time periods and comparisons.
You will need to be mindful of any existing aggregation within the dataset (percentages, ratio, etc.) and how that impacts which measure to select in Auto Insights. More on this to come in part 2 😊.
Default date
Auto Insights sets the first date column in your dataset as the default date. Before you upload data, ensure that the first date column (i.e., the furthest left column) is the most complete date field in your dataset. If the field is null in any rows, those rows will not be counted in Auto Insights.
For example, if you had data around customer service tickets, you could have multiple date columns, such as the ticket created, updated, and resolved date. Since not every ticket would be updated or resolved, using the created date as the default date column would be best.
Transposing data
In the example below, the original dataset has four revenue columns—one for each office and one total column. To get the most out of this data in Auto Insights, the dataset should be transposed to create an office location column to be used as a segment:
Check out the difference this makes in Auto Insights:
With the data transposed, you can simplify the experience, focusing on one measure (revenue, in this case) and get the added benefit of an additional categorical segment field in ‘Office’ to analyze.
Feature Engineering
Below are a few examples of when creating new columns in your dataset can unlock additional analysis opportunities in Auto Insights.
Bucketing measures to create segments
Perhaps you have data with a numerical variable like “age” that has both a unique value and is a descriptor. Before uploading your data to Auto Insights, consider adding another column where you bucket the numerical age data into segments (for example, “25-29”, “30-35”, etc.). Since Auto Insights uses segments to identify drivers of change, this will open up more analysis opportunities. Instead of showing you which specific age occurred the most in the data, these bucketed segments will allow you to understand a fuller perspective and highlight the age groups that contribute significantly to increases or decreases in the measures, reducing risk and increasing reliability in the insights you receive.
Add a ‘1’s column for counts
If you’re interested in understanding how many times something has occurred and you do NOT have a unique ID column in your dataset, you can introduce a new column where each record gets a value of ‘1’ to tabulate the number of occurrences. You can name this column ‘Count’ (or something similar), and this will allow you to have a new measure to analyze where it will sum all the records (the ‘1’ values) to show you how many occurrences are in the specified time periods or dataset. So, if your dataset was sales transactions, you can know the count of how many transactions occurred.
If you have a unique ID column in your dataset, this is an unnecessary step. You can utilize the ‘Number of [unique id column name]’ measure that Auto Insights automatically calculates in order to get the same insight.
Uploading Data
After you’ve made transformations to your data in Alteryx Designer, you can utilize the Auto Insights Uploader Tool to send data directly from Designer and Server to Alteryx Auto Insights. This means you can automate uploads/refreshes and remove risk in the process! Download the tool from the Gallery here.
Once you’ve sent the data to Auto Insights and completed the data setup, it’s worth noting that you have the ability to refactor segment types. You can do this through the Admin Portal by finding the dataset and selecting ‘Edit dataset.’ After clicking through the first 4 screens without making any changes, you get to ‘Confirm Fields’ where you can switch measures to segments (and vice versa). These changes carry through with data refreshes, so you only need to do it once (settings are sticky as long as the columns and names remain consistent across refreshes).
So if you’re not able to see ‘What key changes occurred’ and you know that there are dates in your dataset, you need to into ‘Confirm Fields’ to set your date column type and date format correctly.
Note: Your data will show as a CSV option in the ‘Choose Data Source’ step within dataset admin if it’s uploaded from Designer.
Conclusion
In this blog, we covered best practices around problem formulation and data configuration in Auto Insights. If you're still looking to understand what a good dataset looks like, remember you can use the ‘Show underlying data’ feature on any of the demo datasets within Auto Insights to see an example of an optimal dataset structure to replicate!
If you have any questions, don’t hesitate to leave them in the comments below.
This article is the first in a two-part series on Auto Insights best practices. You can find the second part, which covers Missions and Magic Documents, here.
Additional resources:
Auto Insights Interactive Lessons
6 Steps for Success With Auto Insights
