Cohort • ti

Cohort analysis is the practice of segmenting users or customers into groups (cohorts) based on a shared initial experience, usually a specific time period of acquisition.

By monitoring these cohorts over time, analysts can gain deep insights into retention rates, lifetime value (LTV), feature usage, and behavioral trends that are hidden when looking at aggregate metrics. It shifts the focus from “what is happening now” to “how are groups behaving over their lifecycle”.

The cohort() function in ti provides a tibble and database-friendly implementation of cohort analysis that is compatible with modern backends.

How does it work?

Assigns each member in your dataset to a time-based cohort based on the first occurrence in the .date column
Aggregates distinct counts of cohort members over time
Supports multiple time units (day, week, month, quarter, year)
Returns a segment object, which can be processed using calculate() to generate the cohort table

Arguments

Argument	Description
`.data`	A `tibble` or DBI object containing your dataset.
`.date`	The date column to base the cohort assignment on.
`.value`	The column representing members to track (e.g., customer ID).
`time_unit`	Argument to group your calendar dates: `day`, `week`,`month`,`quarter` or `year`
`period_label`	Whether to use numeric period labels (`TRUE`) or actual dates (`FALSE`). Default is `FALSE`.

Output

The function returns a segment_cohort object. Use calculate() to retrieve the cohort table:

# Example cohort analysis
ti::cohort(.data = contoso::sales,
                  .date = order_date,
                  .value = customer_key,
                  time_unit = "month",
                  period_label = FALSE
             )

This function is basically a database translation of PeerChristensen cohort package