Predictive insights - Insight cards on revenue overview dashboard

This feature will be available to all customers as part of the Extended AI package on the 21st of April.

Insight cards help you understand what’s happening in your KPIs without needing to know which reports to check or what questions to ask.

The system continuously monitors key performance indicators (KPIs) and highlights meaningful changes, anomalies, and opportunities that require attention. Each insight explains what changed, why it matters, and suggests a recommended next step - so you can move quickly from insight to action.

Prerequisites

Availability depends on your subscription and tenant configuration. Extended AI must be enabled as an add-on.

How to use the Insight cards

  1. Open the Engage dashboard
    Navigate to the Engage dashboard. When insights are available, they will appear as cards. After generation, they are also saved under Analytics → Insights for later review.
  2. Review the insight cards
    Select a card to view more details, including a visual graph and recommended actions. Each card explains which KPI changed and why it matters. Take a moment to review the context and potential business impact.
  3. Follow the recommended next step
    Each card includes a suggested action that links you directly to a relevant workflow in Engage. Use this to investigate further, adjust your strategy, or launch a corrective activity.

How it works

Insight generation runs automatically every week during the night between Sunday and Monday. 

The system analyzes your most recent complete week, Sunday to Saturday, and compares the results to historical performance using up to two years of data.

To ensure insights are relevant and actionable, the process follows three main steps:

1. Detecting unusual changes

Machine learning models evaluate the latest KPI results and compare them to expected performance. To determine what is expected, the system considers:

  • Long-term trends, such as steady growth or decline
  • Recurring seasonal patterns, such as holiday peaks

Based on this analysis, the system calculates an expected range for each KPI. An anomaly is defined as a KPI result that falls outside this range and is statistically unusual compared to recent performance. This can include sudden spikes or drops, breaks in established trends, or unexpected behavioral shifts.

Small, random fluctuations are filtered out. Only statistically significant deviations are surfaced as Insight cards.

If no Insight cards appear, common reasons include:

  • No meaningful anomalies were detected
  • Limited KPI data for the analyzed period
  • Missing or incomplete data

2. Explaining the change

When a significant deviation is detected, the system analyzes the KPI in more detail.

Instead of only showing that a number changed, the insight clarifies:

  • How much it changed
  • In which direction
  • How it compares to what was expected

Performance is always evaluated against a relevant baseline that reflects how the KPI typically behaves. This ensures insights are reliable, consistent, and easy to understand.

Each insight card also includes a visual comparison between actual performance and expected performance to provide context.

3. Recommending actions

When the system identifies and explains a meaningful KPI change, it generates a recommended next step designed to help you improve or stabilize the KPI.

Recommendations are created by matching the type of KPI movement, such as:

  • A drop in purchases
  • Increased opt-outs
  • Lower delivery rate

Each recommendation is:

  • KPI-linked
  • Actionable within Engage
  • Routed directly into a relevant execution flow

Over time, the recommendation set expands as more KPIs and patterns are added.

Insights are available in multiple languages and are saved for future reference under Analytics → Insights.

Current KPIs monitored

Currently monitored KPIs include:

  • Weekly contacts who purchased
  • Total identified revenue
  • Average purchase value
  • Delivery rate for email
  • Delivery rate for SMS
  • Unique open rate - email
  • Click-through rate - email
  • Click-to-open rate - email
  • Email unsubscribes
  • SMS unsubscribes

Use-case examples

1. Drop in contacts who purchased

If the percentage of contacts who purchased drops from 4.8% to 3.9% week-over-week, and this decrease is outside the expected range based on historical trends and seasonality, the system detects it as a significant anomaly.

What the insight explains:
The card highlights that fewer contacts converted than expected. It clarifies how large the drop is, compares actual performance to the expected baseline, and explains that the change is statistically unusual—not just normal weekly fluctuation.

Recommended next step:
The insight may suggest reviewing recent campaign performance or creating a recovery or reactivation segment targeting engaged but non-converting contacts. You can move directly into a relevant flow in Engage to take action.

2. Increase in average purchase value

If the average purchase value increases significantly beyond expected levels—outside normal seasonal patterns—the system generates an insight.

What the insight explains:
The card quantifies the uplift and shows how much higher the value is compared to the expected baseline. It clarifies that the increase is not only positive, but also statistically unusual.

Recommended next step:
You may be encouraged to identify the segments or campaigns driving higher basket value and replicate those tactics. For example, you could build a new audience based on high-value purchasers or reinforce successful upsell strategies.

Was this article helpful?

/