This feature is currently in beta, if you have any feedback please contact us.
You can access it by selecting 'Insights' in the left hand sidebar.
Insights provides an easy way to see how your cards are performing. We have developed a simple set of calculated metrics to help you understand how long it takes for customers to interact, to understand the action funnel, and to see whether customer behavior and outcomes differ by device platform.
Watch the video below to learn the basics about sending cards and using the Insights dashboard.
This guide details the reporting metrics presented within the Workbench. It sets out the names, high-level descriptions and business logic for the metrics.
Card activity filters
There are 4 ways to filter analytics:
- Card level. You can select all cards, or a specific card. This includes unpublished cards.
- Platform level. You can select which platform the customer used to interact with their card(s) to see if one is more effective than the other.
- Stream level. You can select
All(all streams combined), the
All cards stream, or any other stream you like.
Selecting All streams will show you metrics across all streams. Selecting the All cards stream will show you metrics for that specific stream.
Example: card A gets sent to the Car loan stream as well as the All cards stream. The customer then views the card in the Car loan container showing the Car loan stream. This will result in having this card 0 views in the All cards stream, but 1 view in All streams.
- Date range, expressed in UTC time. There are many websites that convert your current time to UTC, Time.is is one of them.
How we use cohorts to improve reporting quality
Keep in mind when interpreting analytics charts in Atomic that each data point on the chart represents a cohort of customers. Typically charts show one data point for each day in the selected date-range, which means the data point for that day will be the customers who were sent a card instance on that day.
For example, with the Completion Rate chart, each chart data point tells you how many of the people who were sent a card instance on this day went on to complete that card.
To continue this example: imagine you sent a card on date X. When you select that exact date in the date range, you will then see how many people have interacted with that card between that date, and the current date. If you extend the date range, you will see insights for all people who got sent that card between those dates. Even if they completed the card outside of the selected date range, it will still show in the Insights.
To help you understand how the metrics in our Workbench reports are calculated we've put together the following guide. If you would prefer to work with the raw data and model it with a different approach, or perform analyzes that is more user-centric and less card-centric, or simply join Atomic analytics with other intelligence data you have available, we encourage you to extract the raw data and bring it into your BI and cloud reporting tools.
Analytics data can easily be downloaded from the Analytics Exporter and Analytics Debugger via the Workbench and via API in near real-time and for batch extraction. A full reference of Atomic Analytics Events events is detailed in our API docs.
|Sent||Count||#||Card instances successfully published during the reporting period.|
|Last Sent||Date/time||N units ago||When a card instance from this template was most recently published.|
|View Rate||Calculated||%||Cards which were published and have been viewed at least once within the reporting period.|
|Avg Time to View||Calculated||N units||How long it took for cards to be viewed after being published, for cards which were published and viewed within the reporting period.|
|Completion Rate||Calculated||%||Cards which were viewed at least once and were also completed during the reporting period.|
|Avg Time to Complete||Calculated||N units||Elapsed time from when viewed cards were completed, for cards which were viewed and completed within the reporting period.|
|Action Rate||Calculated||%||Cards which were published and have been interacted with in some way by the user during the reporting period. Excludes API events like cancellation, and expiry.|
|Avg Time to Action||Calculated||N units||How long it took for cards to be actioned after being published, for cards which were published and actioned within the reporting period.|
|Feedback||Count||#||Number of times feedback was given by customers within the reporting period.|
|Snooze Rate||Calculated||%||Cards which were published and snoozed, at least once, by an customer within the reporting period.|
|Links clicked Rate||Calculated||%||Cards which were published and had a link within that card clicked, at least once, by a customer within the reporting period. This data began to be recorded as of June 9th 2023.|
|Avg Time to click a link||Calculated||N units||How long it took for cards to have a link clicked within them after being viewed, for cards which were published within the reporting period.|
|Dismiss Rate||Calculated||%||Cards which were published and dismissed by a customer within the reporting period.|
|Avg Time to Dismiss||Calculated||N units||How long it took for cards to be dismissed after being viewed, for cards which were published and dismissed within the reporting period.|
|Expire Rate||Calculated||%||Cards removed after reaching a predefined expiry date.|
|Expired Unseen Rate||Calculated||%||Cards removed after reaching a predefined expiry date, which had never been seen by a customer.|
|Cancel Rate||Calculated||%||Cards removed via Atomic’s APIs, after being published.|
The insights data is updated from a queue of events, during periods of high activity the insights dashboard may be delayed as this queue is worked through.
^ When filtered, the per-card view rate uses the first
card-displayed event that occurred with the context matching the filters.
Cards that were not displayed in the same platform context as the selected filters, or not displayed at all, both report 0% view rate, as it is true that ‘within the selected context’ view rate was 0%.