Influence overview
11 min
The Influence exploration template in Optimizely Analytics helps you understand how user behaviors, segments, and metrics relate to key outcomes. It highlights the factors most strongly associated with those outcomes, and enables more informed, data-driven decisions. The template operates in three modes based on your selected target: Event, Cohort, or Metric.
As an exploration template, it is reusable; you can save, revisit, share, and integrate configured analyses into dashboards for continuous monitoring and to inform optimization efforts. It helps you:
- Identify key drivers – Uncover which specific user events, metrics, or cohorts are most strongly associated with your key business outcomes, such as revenue, retention, or conversions.
- Understand impact and prioritize – Gain clarity on the magnitude of influence these factors have, letting you focus on changes that yield the greatest impact.
- Make data-driven decisions – Move away from guesswork with a systematic, data-backed approach to decide what to goal against, what to test next, or what features to build.
- Bridge the gap – Democratize advanced analytical insights, letting PMs, analysts, and experimentation teams to explore relationships without needing specialized statistical tools.
Important
Before you create explorations inside Optimizely Analytics, create datasets and connect your data sourcesconnect your data sources.
How does the Influence template work?
The Influence template is a structured framework designed to guide you through a systematic exploration of your data. Each module plays a crucial role in building your exploration, leading you from a broad question to actionable insights. Within the template, you must configure the following modules:
Configure your analysis using the following modules:
- Define your core question (Target and Test modules)
- Target module – Define the target, which represents your primary outcome of interest. This could be a specific event, a key metric, or a cohort. This selection sets the stage for the entire analysis and determines the type of relationships you will explore.
- Test module – Define the tests, which represent your potential influencing factors. These are the events, metrics, or cohorts you think might drive or associate with your target. You must select at least one test. The combination of your target and test types (for example, Event → Event, Metric → Cohort) forms the fundamental question the template will answer.
- Refine the scope of your analysis (Actors and Time range modules)
- Actors module – Define which actors are included for your analysis. You must specify the primary dataset (for example, Users, Accounts) and can optionally filter this population by specific cohorts (for example, First Time Visitors).
- Time range – Define the overall time period for your analysis. For Event → Event analyses, you can further refine this with a Max time between test and target window to specify the temporal proximity required for an influence to be considered.
- Interpret your results
- Statistical interpretation – Have the relevant statistics (for example, Likelihood Impact, Metric Impact, Correlation) calculated automatically when your target, test, actors, and time range are defined.
- Visualization module (Optional) – Visualize your results using different types of charts and apply filters or sorting to highlight key findings.
- Filters module (Optional) – Dynamically narrow down the data presented in your visualizations without altering the core question. It is ideal for exploring "what if" scenarios or examining subsets of your data.
Consider your target and test types
The Influence template helps you answer specific questions by defining your target (the outcome you want to understand) and your test (the potential influencing factor you want to investigate). To choose the appropriate target and test options, consider your primary analytical goal:
Goal 1: Understand what drives a specific user action or conversion
The following scenarios demonstrate this goal:
- An online commerce site can determine if viewing a pricing page leads to a purchase.
- A streaming service can analyze if watching a trailer influences a user to play content.
For this,
- Choose Event as Target. This lets you identify which factors are most influential in leading to that critical event.
- Choose your Test type based on the influencing factor you want to investigate:
- Event as test – If a specific user action (for example, Add to cart, View pricing page) might influence your target event.
- Metric as test – If a metric (for example, Average session length, Number of interactions) might influence your target event.
- Cohort as test – If a cohort (for example, First time visitors, Mobile users) might influence your target event.
Goal 2: Optimize a metric
The following scenarios demonstrate this goal:
- An ecommerce platform can determine if using a discount code increases Revenue per user.
- A software company can analyze if participating in a webinar improves the Retention rate.
For this,
- Choose Metric as Target. This lets you identify factors strongly correlated with or impacting the value of that KPI.
- Choose your Test type based on the influencing factor you want to investigate:
- Event as test – If a user action (for example, Used feature X, Completed onboarding) might influence your target metric.
- Metric as test – If another metric (for example, Number of items viewed, Session duration) might correlate with your target metric.
- Cohort as test – If a cohort (for example, High-value customers, Power users) might impact your target metric.
Goal 3: Understand, define, or grow a specific cohort
The following scenarios demonstrate this goal:
- A retailer can determine if making more than three purchases a month identifies someone as a High-value customer.
- A subscription service can evaluate if a significant drop in login frequency predicts a user entering the At-risk churn cohort.
For this,
- Choose Cohort as target. This lets you identify the factors that predict or define belonging to that target segment.
- Choose your Test type based on the influencing factor you want to investigate:
- Event as test – If a user action (for example, Completed survey, Visited X pages) might predict membership in your target cohort.
- Metric as test – If a metric (for example, Lifetime value score, Product usage frequency) might predict membership in your target cohort.
- Cohort as test – If another cohort (for example, Trial users, Loyalty program members) might be associated with your target cohort.
Build an influence analysis
When you determine your goals, you can select the appropriate target types and test types and build your analysis.
- Click + > All Exploration templates > Influence.
- Name your analysis, add a description, and choose the folder you want to place it in.
- Select a dataset from the Actor field in the Actors module, to define the primary unit of analysis (for example, Users or Accounts). Then, optionally select a cohort in the Included Actors field to narrow down your analysis to a specific group. The default is set to All Users.
- Configure the Test module:
- Select a type from the Type field. The available types are Event, Cohort, and Metric. For information on each type, see Test types.
- Select the events, cohort, or metric you want to use as your test variable, based on the type you selected.
- (Optional) Select an attribute in the Group by Test Property field to group your test event by a specific property in the selected dataset.
- Configure the Target module:
- Select a type from the Type field. The available types are Event, Cohort, and Metric. For information on each type, see Target types.
- Select the event, cohort, or metric you want to set as your target, based on the type you selected.
- Select a Direction if both your test and target are events. The following are the two options:
- Causes (Before Target) – Test events that occur before the target event.
- Effects (After Target) – Test events that occur after the target event.
- Set the Max time between test and target to define the time window for your analysis (for example, 7 days).
- Preview your configuration in the Influence Timeline and click Run.
- View the results plotted in a graph in the visualization window.
Example: Analyzing the likelihood of users playing content after viewing similar content
The following example creates an influence analysis that shows the likelihood of users who perform the View Similar event to perform the Play Content event in a seven-day window. For detailed information on individual elements, see Understand your Influence analysis.
- Click + > All Exploration templates > Influence.
- Name your analysis, add a description, and choose the folder you want to place it in.
- Select the Users dataset from the drop-down list in the Actor field.
- Select the First Time Visitors cohort in the Included Actors field. This field is useful if you have a large number of users and want to narrow down your analysis to a selected group. The default is set to All Users.
- Configure the Test event.
- Select Event as the test type using the Type drop-down list. There are three types: Event, Cohort, and Metric.
- Select the View Similar event using the Events drop-down list.
- Select an attribute in the Group by Test Property field. This lets you group your test event by a specific property in the selected dataset.
- Configure the Target event.
- Select Event as the test type using the Type drop-down list. There are three types: Event, Cohort, and Metric.
- Select the test events, cohort, or metric based on the type you choose.
- Select Causes (Before Target) as the Direction for the target event. There are two options for direction:
- Causes (Before Target) – Test events that occur before the target event.
- Effects (After Target) – Test events that occur after the target event.
- Set the Max time between test and target to 7 days.
- Preview your configuration in the Influence Timeline and click Run.
- View the results plotted in a graph in the visualization window.
- The X–axis shows the percentage of all the included actors who performed the test event at least once.
- The Y-axis represents the Likelihood Impact, showing how much more likely users who performed the test event are to also perform the target event, compared to all users. For example, a +50% value means these users are 50% more likely to perform the target event.
You can set multiple filters inside an influence analysis. Whenever you create filters, you can reference parameters inside them to fine-tune your exploration. See Parameters overview for information.