Predictive Modeling: Data Requirements

Introduction: Leadspace Predictive Scoring

Predictive lead scoring with Leadspace takes the traditional lead scoring model and optimizes it, taking it to the next level through the application of big data and machine learning algorithms. At Leadspace, we evaluate the key facets of existing customers, prospects, and opportunities—ranking them against a scale that can distinguish customers and prospects who are more likely to convert, retain, or buy from the company’s products and services. 

To learn more about Lead Scoring and Predictive Modeling, refer to our Lead Scoring: A Primer article.

Predictive scoring works like this: 

1. Analyze a large number of past deals to find the signals in the noise that positively correlate with closed-won deals. (At Leadspace, we do this with a proprietary scoring engine we’ve built and enhanced over the years, crunching thousands of signals to separate the merely correlative from the causal.) 
2. Based on those identified signals, create a model of what an ideal buyer looks like. 
3. Test your model against a lot of closed-won deals to validate the match.
4. Run incoming leads through the model to apply scores.
5. Route or prioritize the leads with the highest scores. Recycle or nurture those with lower scores.

Getting Started with Predictive

As mentioned, Leadspace approaches predictive by analyzing large quantities of data at scale to build a model that comports with your business requirements. 

What do we need to get started?

As a part of our modeling process, we join, enrich, and deduplicate disparate data objects to build a unified representation of your historical sales funnel. For any predictive project, Leadspace requires three (3) files:

1. Leads file
2. Opportunities file
3. Accounts / Contacts file (depends on the model type)

Required Data for Predictive Modeling

Data Required for all 3 Files:

Unique Identifiers

• Unique ID Field: Mandatory
  • Must share a Unique ID field that will be the basis of joining and merging the files.
    • Account model: all files should have an Account ID field
    • Lead Model: all files should have a Lead / Contact ID

Date Fields

• Creation Date: Mandatory
  • All files should include Created Date field, allowing us to understand timeframes across the funnel

Look back Window & Number of Records

Leadspace generally will gather as much data on your Leads, Opportunities, and Accounts as possible. However, if your Go-To-Market Strategy has changed significantly in the last two (2) years, we'd recommend sharing all data from the date of that strategic change - and forward. 

• Recommended Look back Window: 24 Months
  • Updated data from the past 24 months - generally, this would represent your current customers
• Number of Records: Minimum: 50K 
  • Minimum: 50K unique records in total (all 3 files).
  • Preferred as many records as possible in the Opportunities file (most important file).
• Conversion Rate: 0.5-10% Positive Conversion
  • Include as many success records as possible from all the data (preferred closer as much as possible to 10%).

Opportunities file requirements:

• Conversion Definitions: Mandatory
  • Must have a clear definition of a positive conversion
  • The definition of a positive conversion varies across companies and verticals, so we'll need to receive definitions of your stages. However, a good rule of thumb for a positive definition is Stage = "Closed Won").
• Opportunity End Date: Mandatory
  • Must contain an End Date filed.
  • This allows us to understand when an opportunity was considered Closed Won / Closed Lost
• Aggregation Fields: Recommended
  • Most modeling projects contain an aggregation value field (used for the calculation of the revenue projections).
  • If there is more than one revenue field (e.g. Gross ASV, Net ASV), we’ll need customer clarification as well.

Mandatory and Optional Fields for Leads, Opportunities & Accounts