Overview of Infer for Salesforce

Infer for Salesforce provides advanced predictive scoring integrated directly into Salesforce.com. When a new Lead is created in Salesforce, or an existing Lead is modified, Infer transparently detects it, calculates a score, and pushes it back to Salesforce where it can be read on the Lead object itself.

While most often Infer Scores are attached to Lead Objects, in some cases Infer Scores may be attached to other Salesforce Objects such as Contacts, Accounts and Opportunities, in addition to or instead of Leads.

Because of the transparent nature of Infer’s integration with Salesforce, only minimal administrative support is needed to set up and operate the solution, specifically:

  • Granting Infer appropriate access to Salesforce
  • Installing and configuring the Infer AppExchange package
  • Adding Infer Score and/or Rating fields to Salesforce Views.

 

The Model Building Process

An Infer Score is called predictive because it is based on a predictive model. Each customer’s unique model is built by applying machine learning to historical information from their Salesforce instance.

⇒ Your Infer Model must be built before Scores can be pushed into Salesforce

Infer handles all aspects of the model development. However, because the predictive
technology learns from historical outcome information, Infer needs access to Salesforce in order to build the model.

⇒ To build the model, Infer needs a Salesforce account with Access All Data privileges.

See Setting up Infer for Salesforce for details on granting Infer appropriate permissions. Once Salesforce access is set up, Infer builds the model with a three step process: Capturing historical outcome information, extending the data with thousands of signals from Infer’s data library, and applying machine learning algorithms to extract the signals that help predict positive outcomes.

From time ­to ­time it is desirable to Rebuild the Infer model. Rebuilding allows the model to learn from newer outcome information, as well as being able to take advantage of new external data
within Infer’s data library.

⇒ Rebuilding a model and pushing it to production can be done entirely by Infer.

If you are making major changes to your sales processes or the data in Salesforce, please contact Infer to discuss whether the model should be adjusted or rebuilt to reflect the changes.

⇒ Pushing a new model to production involves both using the updated model to score new Leads and rescoring the backlog.

Typically, new models are normalized to align with the previous one, which means that the mapping of scores to ratings does not need to change. If the model changes are substantial, it is possible that aligning the normalizations will not be desirable, in which case the rollout of the new model will be coordinated with the appropriate configuration change.

 

Scores and Ratings

Infer Scores are always normalized to be between 0 and 100 and can be read from a field called Infer Score directly on a Lead or other object. The Score should be understood as an abstract representation of the quality of the lead, so that Leads with higher scores are of higher quality than Leads with lower scores.

To simplify the interpretation of scores, it is common to map scores into ratings. One example would be: “A” leads are those with scores between 80 and 100, “B” leads have scores between 50 and 80, “C” leads score between 30 and 50, and “D” leads have scores below 30. The Infer AppExchange package automates the mapping of scores to ratings with configuration options for the names of the rating buckets and the cutoffs.

The Infer Score and Infer Rating fields can be added to Salesforce Views as with any custom fields. See the section Adding Infer Fields to Salesforce Views for details.

⇒ Typical best practice is for sales reps to see the Rating but not the score.

 

Live Scoring and Backlog Scoring

Live Scoring refers to attaching scores to Leads or other objects as they are created or modified. Backlog Scoring is the process of attaching scores to objects that existed prior to setting up Infer for Salesforce. Attaching scores to pre­existing leads enables a variety of important use cases, notably identifying high quality leads that should be promoted out of a nurture status.

After the Infer Model is completed, Live Scoring will begin and Backlog Scoring initiated. Typically, backlog scores will be first calculated and then pushed to Salesforce as a batch.

⇒ The time it takes to score the backlog depends on the number of Leads to be scored.

If your backlog is very large, contact Infer for an estimate of the time needed or backlog scoring to complete.

Backlog scoring works the same for all object types ­­ Leads, Contacts, Accounts, Opportunities with one exception.

⇒ Because of Salesforce limitations, scores cannot be attached to already
converted Leads.

This limitation must be kept in mind when reviewing reports over Leads that extend back prior to going Live with Infer. See The Infer Dashboards for details.

 

The Infer AppExchange Package

The Infer AppExchange package is easy to install and automates all of setup of the Infer for Salesforce solution apart from granting Infer access to Salesforce.

The Infer AppExchange package includes the following components:

  • Custom fields on the Lead, Contact, Account and Opportunity objects to hold the Infer Score and the Infer Rating 

Exposing Infer Score or Infer Rating fields in Salesforce Views is handled in the same way as any Salesforce custom fields. See Adding Infer Scores to Salesforce Views for details

 

  • A set Salesforce Dashboards that make it easy to use the score for a variety of use cases. See The Infer Dashboards for details

 

⇒ The Dashboards are available only for Leads

 

  • A configuration page that controls
    • Which objects should be scored
    • The mapping from Scores to Ratings (e.g. A,B,C,D)
    • A few preferences used in generating the reports

 

 

Technical Considerations

When using Infer for Salesforce, a custom Score object will be created for every object of the type(s) that are to be scored, Lead, Contact, Account or Opportunity. The Infer Score ID field on the core object (Lead, etc.) is a reference field pointing at to the corresponding Score object.

This setup has several implications that should be noted.

  • Storage Requirements. Each custom object is counted as 2Kb of storage in Salesforce. One such custom object will be needed for every object that gets scored by Infer. 

 

Make sure that the storage required for the scoring is available in your
Salesforce Org.

 

  • Mapping Lead fields on Conversion. The Infer Score fields on Leads are formula fields and therefore cannot be mapped to to Contacts, Accounts or Opportunities upon Conversion using the Salesforce “Map Lead Fields” feature. 

 

⇒ If you need scores on objects other than Leads, contact Infer to set up mapping of scores or direct scoring of these objects.

 

  • Triggers. It can be convenient to define a trigger to fire when an Infer Score is written or updated. However, because Infer writes the score to the Score object, rather than the directly to the Lead or other standard object, the trigger must be defined on the Score object. See the section Defining Triggers on Infer Scores for details.

 

  • Validation Rules. If an object are not compliant with Salesforce validation rules, it may be impossible for Infer to attach a score. For this reason, you may want to exempt the Infer User from validation rules. See Backlog Scoring and Validation Errors for details.
Was this article helpful?
0 out of 0 found this helpful
Have more questions? Submit a request

Comments

Powered by Zendesk