Skip to main content

Open Model Hub

Objectiv comes with an open model hub that contains a growing collection of open-source, free to use data models that you can take, stack and run to quickly build highly specific model stacks for product analysis and exploration. It includes models for a wide range of typical product analytics use cases.

All models are compatible with datasets that have been validated against the open analytics taxonomy. The source is available for all models and you’re free to make any changes. You can use the included pandas-compatible Bach modeling library to customize them, or even add in advanced ML models.

Input and output of the models are in the Objectiv Bach DataFrame format, which enables you to chain multiple models together.

Models

Each model can run before, after or combined with any of the available other models. This allows you stack them for your analysis and further ML.

first models in hub

We just released a first version of the open model hub to show the high-level working. Each of the models below is now in development and will be added during the next weeks.

Users

  • unique_users: Unique users.
  • user clustering: Grouping uses based on similar behavior in using product features.

Product feature usage

  • features_per_user: Amount of users for each product feature.
  • product feature scoring: Assign scores to product features that define how they help users reach a goal.

Conversion

  • conversion: Unique users that reach a defined conversion feature.
  • conversion_funnel: Sequence of events before a conversion.
  • marketing conversion funnel: Comparing user conversion funnels between marketing campaigns.

Retention

  • retention: Percentage of users retained per cohort.
  • frequency: Users categorized by number of total sessions.
  • recency: Mean number of days before users return.

Sessions

  • sessions: Unique sessions.
  • session_duration: Mean session duration.
  • duration_between_events: Mean session duration between specific defined events.
  • bounce_rate: Sessions with only a single hit.

Available filters

Filters can be applied to each of the models. Initially, there is a first filter to quickly zoom into new users or sessions.

  • is_first_session: The first-ever session of a user.

Model parameters

Each of the available models can run on a desired set of common parameters that define:

  • time range
  • time aggregation
  • user app
  • feature selection

Each model also has specific parameters that can be customized, for example which event defines conversion.

Output to BI

Once you are done building a model, you can output it to an open-source BI tool with a single command. Right now, we support Metabase, with more to come.

Demo

see the open model hub in action

You can easily run the full Objectiv pipeline in a docker demo, see the quickstart guide.