An Insight is built using one or more of three elements:
- Reports and their results
- Report plugins — uses a report as data input. Examples are timeseries forecasting, A/B testing or anomaly detection.
- Intelligence plugins — uses a model with features as input. Examples are predictions, clustering or recommenders.
Most Insights combine a few of those to bring end-users fully automated insights into their data. An example would be an Insight that gives details into the retention of users of a website or app:
- It could first show how the results for the last 90 days using a simple retention report per campaign.
- After that it can run and display the results of a forecast retention report plugin to show how the next 14 days are predicted in terms of retention per campaign.
- Finally it can use an associative-rules based intelligence plugin to present what user behavior leads to higher retention.
The next sections will describe each in more detail with code examples.