Predictive Insights - Model Building Guide

This guide is a basic step-by-step tutorial on building models in the platform. For a comprehensive overview of our modeling process, please see the Predictive InsightsModeling Guide.

Create New Market

Picture1.png

Navigate to the “Markets” icon on the left-hand side and select the “create” button to build a new model.

 

 

 

 

Picture2.jpgSelect the type of model that you would like to build. Single product models will incorporate only one product or service while multiproduct models will incorporate multiple products or services.

 

 

 

Picture3.jpg

 

 

Next, select the type of records that you will be using in your model. Lead-based models are comprised of individual-level records and account-based models are made up of organizations.

 

 

 

 

 

Upload Data Files

Upload a CSV for the positive set and a CSV for the house list. See our data requirements below for a full overview of the data required to build a model.

Positive List

The positive set is a subset of the total universe, where the desired outcome has occurred. A positive set should contain at least 400-600 unique records. For example, if we are optimizing for closed-won, we take the closed-won records from the CRM. If an organization is short on data, they may wish to include late-stage opportunity records in the positive set as well as closed-won records.

House List

The House List is the universe of potential targets (prospects), in your systems, for this model. A house list should contain 10 times the number of records in your positive set. If your positive set had 600 records, your house list should contain 6,000. If you are trying to optimize the Marketing funnel, then the relevant universe consists of all of the leads in your marketing platforms. If you are trying to optimize the BDR performance, then the relevant universe consists of your MQLs that are passed on to BDRs.

Map the Fields

The final step in the model creation process is to map the fields between your files and the Predictive Insights fields. Click the drop-down menus to match the field names on the left with the corresponding fields within the Positive and House sets. In the example below, you can see the Company Name field has been selected from the dropdowns to match with the Predictive Insights Company name.

All fields in the Required & Recommended Fields column with an asterisk need to be mapped in order to move forward with the model.

Picture4.png

 

The last step is to click “Create” to initiate the building of the model. You will see the message below in the bottom-right corner of your screen, and you will receive an email once your model is ready.

Picture5.png

 

 

 

 

 

 

Data Requirements

 Positive Set: 

  • At least 400-600 records, if you have more, please include them.
  • Typically closed/won opportunities or opportunities.

House List:

  • x10 records compared to the positive set.
  • Typically, non-converted leads.

Time Frame:

  • Past 18 months, if possible.

Account-Based Model Data Requirements

The basic data fields for account-based models are as follows:

  • Company (Account) Name - Required.
  • Company Status/Stage - Required.
  • Company Domain - Not required but good to have.
  • Email Address - Not required but good to have (email addresses of leads that are associated to the company).
  • Company State and Company Country - Not required but good to have.
  • Creation Date - Not required but if not included please provide time range.
  • Any other available fields.

Lead-Based Model Data Requirements

The basic data fields for lead-based models are as follows:

  • First/Last Name (in one field or split into two fields) – Required.
  • Email Address – Required.
  • Company Name – Required.
  • Title – Required.
  • Source – Not required but good to have.
  • Stage: Lead anything pre-MQL, MQL, MTG CREATE, SQL, SAL, Closed/Won.
  • Company Domain - Not required but good to have.
  • Creation Date - Not required but.

In case of a closed/won optimization use case, positive records provided should be those associated with the closed/won activity. Ideally, they should be the contact you started and completed the opportunity conversation with (the record that came through the lead flow).

Collecting Data for Common Use Cases

Use Case

Details

House List

Positive Set

 

We would like to score our inbound marketing leads and determine which ones are likely to convert

Marketing automation database leads

Marketing leads converted to opportunities during the last year

Outbound ABM account sourcing

We would like to identify good accounts for an outbound ABM multi-touch activity

Marketing automation database leads

Marketing leads converted to opportunities or closed won accounts during the last year

BDR queue prioritization

We would like to optimize the BDR conversion rate from MQLs to sales opportunities

MQL records from the last year

Leads converted into opportunities by BDRs over the last year

Product line scoring - Leads

Which of our prospects are likely to buy a specific product?

Contacts / Leads from Salesforce

Closed won leads for this product

Product line scoring - Accounts

Which of our prospect accounts is likely to buy a specific product?

Accounts from Salesforce

Closed won accounts for this product

High-value customer prediction

Which of our prospects are likely to purchase at a higher customer value?

Accounts / Leads / Contacts from Salesforce

Closed won accounts / leads with above ASP $ value

0 Kudos
Contributors
Latest Articles
Tuesday
2 weeks ago
2 weeks ago
Labels (1)