Pipeline Predict Scores indicate how likely an account is to generate a new pipeline opportunity in the near future. Using machine learning, Demandbase analyzes historical opportunities, account characteristics, and engagement activity to identify patterns that predict future opportunities.
These scores help you prioritize accounts for sales outreach and identify potential renewal, upsell, and cross-sell opportunities with existing customers.
To configure your Pipeline Predict Scores, see Set Up Pipeline Predict Scores.
Tip: You can configure multiple Pipeline Predict Scores for different product offerings. See Understanding Multiple Predictive Scores.
Quick Facts about Pipeline Predict Scores
| Score |
|
| Initial Model Build Time | 24 hours. |
| Model Re-Training Time | 24 hours. See Retrain Pipeline Predict Scores. |
| Score Update | Automatically updated nightly. |
| Customer Connection Modes |
|
Data Used to Calculate Pipeline Predict Scores
Demandbase uses the following data to calculate Pipeline Predict Scores.
Important: The Pipeline Predict model only uses activities assigned Engagement Points greater than 0. Make sure you assign Engagement Points to relevant, high-value activities you want included in the model.
Data |
Description |
|---|---|
| Past Opportunities |
New business opportunities opened within the past 12 months, from at least 50 distinct accounts. See Opportunity Criteria for Pipeline Predict. Tip: By default, the model uses opportunities with a Created Date within the past 12 months. |
| All Activity Logged in CRM (account/contacts/leads) | Activity type, category, and role for activities logged against accounts, contacts, and leads. Activity details are not used. |
| Engagement Points |
Activity type, category, and role. Activity details are not used. Important: The model determines the relevant time window for activities, including how recently each activity occurred. Don't use Activity Date filters in the Engagement activities selector. |
| Sales Inbox/Calendar Events | Email opens, clicks, and calendar events. Email subjects and event details are not used. |
| MAS: All email activity | Email opens and clicks. Email subjects and content are not used. |
| Firmographics | Country, state, industry, revenue range, and number of employees. |
| Trending Intent (daily update) | Daily intent signals. |
| Historical Intent | Intent behavior up to one year before the opportunity Created Date. |
| Known Website Visitors | Visitor job roles and number of visits. Specific pages visited are not used. |
| Unknown Website Visitors | Number of unknown page visits. Specific pages visited are not used. |
Example Pipeline Predict Score for an Account
The following example shows how Pipeline Predict evaluates account data to calculate a score.
Opportunity Criteria for Pipeline Predict Model
The Pipeline Predict model requires opportunities that meet both of the following criteria:
New business opportunities opened by at least 50 distinct accounts. See New Business Opportunity Definition in Pipeline Predict Model.
-
Opportunities opened within the past 12 months.
Tip: You can configure a date range shorter than 12 months. However, we recommend using the full 12-month period to include enough historical data for accurate predictions.
How Pipeline Predict Identifies New Business Opportunities
Pipeline Predict considers an opportunity to be new business if the account was not in a Customer designated Journey stage when the opportunity was created.
Important: The model uses the stage designations configured in the Global Journey Funnel.
An account is considered to be in a Customer stage if either of the following conditions applies:
The Journey Stage is explicitly designated as Customer.
- If the Journey Stage isn't explicitly designated, the model evaluates its name using the following criteria:
- Includes at least one of these keywords: customer, expansion, closed/won, existing, upsell, renewal, client, or churn risk.
- Does not include any of these keywords: lost, not customer, ideal customer profile, or ideal client profile.