We’ve compiled this list of tips to help you when using the Data Analyzer.
Beta This is a beta feature, which means we’re still developing it. Some functionality might change.
Since Data Analyzer builds models around an exact value, a good way to build models on numeric fields is to use a cut-off or categorization formula field:
-
Cut-off - You should create a formula field that calculates if the numeric field is more/less than a benchmark that you set. Then run the Data Analyzer to gain insights for your business case being above or below the benchmark.
Example: I would like to gain insights on what is driving the margin percentage on my projects. From my business acumen, I know 25% is a good benchmark/cutoff, so I created a checkbox formula field that checks if the project had more/less 25% margin. -
Categorization - You should create a formula field that categorizes values from a numeric field. Then run the Data Analyzer to gain insights on what are the main factors behind each category.
Example: I would like to gain insights on what is driving my customers’ satisfaction score. I am currently capturing these with a score between 1 and 10. I can make 3 categories based on the score - Low (1-3), Medium (4-6), High (7-10). Then I can run the Data Analyzer once for each category to understand what factors are driving each one.- Use the power of Quickbase reports when using the Data Analyzer to exclude fields and records that you don’t want to be evaluated, especially for data that is contextually the same as the outcome.
Example: If I am evaluating the categorized satisfaction score of customers, I would remove the original satisfaction score from the data. - Key information might be in a different table. Summary and Lookup fields are crucial in that case. If you want the AI to evaluate as best as possible, add all the information that you think might be relevant for the desired outcome.
- Date and datetime fields are not analyzed, but the AI is checking if factors are becoming more significant over time.
- Fields containing a lot of information (i.e., Multi-line text, Comments, Address) are very likely to be ignored as the information in the fields is not standardized. You can however extract specific information from it (e.g., Comment field contains word “Closed”, “State” from Address field)
- Use the power of Quickbase reports when using the Data Analyzer to exclude fields and records that you don’t want to be evaluated, especially for data that is contextually the same as the outcome.
FAQs
Q: What data can I analyze and get insights on?
A: Your data needs to meet a minimum set of requirements to work with Data Analyzer. However, as long as the data and outcome are in the same table (or table report) in Quickbase, Data Analyzer will look for statistical trends in them and build a model. You can check the sample data in this app to see how Data Analyzer can solve various problems. We don't recommend sending sensitive or personal data to Data Analyzer as it result in biased insights or be recorded in the template application.
Q: What are the limitations of Data Analyzer?
A: Data Analyzer does not analyze DateTime fields, however, it does check for factors becoming more or less significant over time. The Outcome field can be a Checkbox, Multi-select, or Text field. Data Analyzer won't look for potential factors outside of the report/table defined in the wizard—any factor that could be contributing to the outcome needs to be added to the report/table being analyzed.
Q: Will Data Analyzer make changes to my data or degrade the performance of the app?
A: No. Data Analyzer doesn't make any changes to analyzed data. Deploying the prediction formula is a manual operation. It's highly unlikely for Data Analyzer to affect the performance of an app.
The Machine Learning (ML) Model
Q: Are you training your AI models with my data?
A: No. Data Analyzer utilizes Machine Learning technologies. The AI (ML) model is stored and presented to you in the template app only. It's trained on statistical trends in your data that correlate to the desired outcome without any contextual analysis.
Q: How to create a new ML model?
A: After performing the 1-time installation of the plugin, you can create a new model. In it, you need to set where the data you want analyzed is in Quickbase and what is the desired outcome that you want to predict based on that data.
Q: How long does it take to create a model?
A: It depends on the amount of data you select for analysis. It could be anywhere from a few seconds up to more than 15 minutes.
Q: What is the prediction formula?
A: The prediction formula is a representation of the ML Model, which is based on logistic regression. It calculates the probability for the outcome to occur based on trends in the data.
Q: Why is the prediction formula valuable?
A: By deploying the formula to your original application and data, you can get live predictions for the outcome in both new and existing records. This could allow you to make proactive decisions.
Q: How often should I use Data Analyzer on the same outcome?
A: General rule of thumb is every 1-2 months. As the ML model is static at a point in time, creating a new one to include newer data is crucial for an accurate prediction. Newer data can include both new records and new fields that are potential factors for the outcome.
Hint: You can check the “Deployed” checkbox in your model and you will receive an email 2 months after the model creation date.
Data Analyzer Template App and Plugin
Q: Can I make changes to the app?
A: You can use the app to create new models and store the results of your analysis. We don't recommend editing the existing app schema (tables/fields) as it may prevent Data Analyzer from working correctly.
Q: Can I use the template app like my other Quickbase apps?
A: Yes. Apart from the provided schema, reports, forms, etc., the template app is the same as a standard Quickbase app. You can assign roles, build tables and forms, create records, build pipelines, set up notifications, etc.
Q: How many Data Analyzer Template Apps/Plugins can I have in my realm?
A: As many as you'd like. However, note that the plugin should be installed only once per template app, so whenever you open an existing Template App, you should check if the Data Analyzer Plugin is installed before attempting to run the installation. This can be checked in Installed plugins.
Q: Why can't I install the Data Analyzer plugin?
A: There are 4 requirements for installing the Data Analyzer plugin.
- You are on a Business or Enterprise plan.
- Plugins are allowed to be installed by App Admins in the realm (Admin Console → "Permissions").
- You are installing the plugin in the Data Analyzer Template App.
- You are an App Admin in the app.