The Pipeline Predict Score tells you how likely it is that an account will become a pipeline opportunity. This insight tells you which accounts are ripe for targeting, provides insights into pipeline planning, and is a crucial insight in determining when an account is ready for sales outreach.
Click the links below to learn more:
- How Does It Work?
- Quick Facts
- Factors Used When Calculating Pipeline Predict Score
- Accessing Pipeline Predict Score Results
- Take Action with Recommended Accounts
How Does It Work?
Pipeline Predict is a machine learning model that looks at past opportunities from your Customer Relationship Management system (CRM), learns account properties and activity patterns that are leading indicators of new CRM opportunities being created at accounts, and then identifies those leading indicators in other accounts.
- Model Build Time: 24 Hours
- Score: 0 – 100% (A score of 75+ is worthy of nurturing and 90+ requires immediate attention.)
- Initial Model Build Time: 24 hours (The model starts running when you click the Save and train model button in Settings > Demandbase-Wide Settings > Predictive Score Setup.)
- Model Re-Training Time: 24 hours (The model starts running when you click the Save and train model button in Settings > Demandbase-Wide Settings > Predictive Score Setup.)
- Score Update: Automatic (nightly)
- Customer Connection Modes:
- For those customers who upload a .csv file the scores use the following data: Creation Date (<13 months), Account Website and Owner fields.
- Connected Mode (fully integrated with a MAS and a CRM system): Full scoring capability.
- Hybrid Connection Mode (integrated with a MAS or a CRM system): Are required to upload open opportunities to obtain results from the scoring model.
- Disconnected Mode (do not have a CRM or MAS integration): Are required to upload a .csv file with open opportunities to obtain results from the scoring model.
Factors Used When Calculating Pipeline Predict Score
All available data is used, including:
- Any set of 50 or more. For best results, we recommend using 100 or more.
- Past Opportunities from CRM
- Any opportunity created in the past six months (these can still be open, closed/won or even closed/lost) for net new customers are used as default positive examples to train the Pipeline Predict Model.
- You can also use Selectors to select the opportunities as positive examples to train the Pipeline Predict Model.
- We are not excluding all closed/won from training data, only those that are from more than six months ago.
- All Activity Logged in CRM (account/contacts/leads)
- Pipeline Predict uses activity and role type. It does not use activities details.
- Sales Inbox/Calendar Events
- Pipeline Predict uses email open and click and the existence of calendar events. It does not use the title or details in email/calendar events.
- MAS: All email activity
- Pipeline Predict uses email open and click, not including the title and content.
- Pipeline Predict uses country, industry, revenue range and number of employees.
- Trending Intent (daily update)
- Accounts will be predicted using daily intent signals.
- Historical Intent
- Demandbase looks at the intent behaviors 30 days prior to the opportunity creation date to inform/train the model.
- Known Website Visitors
- Pipeline Predict uses which job role visits the website and the number of visits for that job role. It doesn’t use what pages were visited.
- Unknown Website Visitors
- Pipeline Predict uses the number of page visits and doesn't use what pages were visited.
The image below shows an example of how the Pipeline Predict Score is calculated for an account.
Top Factors that Determine the Score
The example below shows some of the top factors that determine the Pipeline Predict Score across all accounts. Each account will have different “top factors” that determine its score.
Accessing Pipeline Predict Score Results
After an Administrator configures the Pipeline Predict Score, you can access Pipeline Predict score results. From the left navigation bar, go to Analytics > Dashboard and select an account list from the drop-down list. Scroll down to the Accounts section and view the Pipeline Predict tab. This tab shows the accounts recommended using the Pipeline Predict Model.
Take Action with Recommended Accounts
You can take action to engage the recommended accounts. For more information, see Taking Action with Accounts and People.
- From the Pipeline Predict tab, click See all accounts.
- On the Engagement page, scroll down to the table and click the Accounts tab.
- Select the accounts you want to take action on and click Take Action.
- Select the action you want to take for the selected accounts. For example, you can create a task in Salesforce to have these accounts assigned to a person on your Sales Team for follow up.
Do the scores work out of the box?
Yes, they are available 24 hours after your CRM and Marketing Automation System (MAS) integrations are enabled and the user clicks the Save and train model button on the Predictive Score Setup page. At least one Keyword Set needs to be created.
To access this page, from the left navigation bar, go to Settings > Demandbase Wide-Settings > Predictive Score Setup.
How long does it take the model to run?
How does Pipeline Predict improve over time?
The model will automatically pick up new signals every night from all available sources.
What’s a good score?
75+ is worthy of nurturing and 90+ requires immediate attention.
How does the Pipeline Predict Score know which Personas/Titles are more relevant?
The model will decide automatically based on all the data plus the “level” of the title (CMO → VP → Director → Manager).
How do I see what factors influence the Pipeline Predict Score for a specific account?
On the Pipeline Predict tab, click the account you want to view. In the Pipeline Predict Score card on the Dashboard tab, hover over the labels (e.g. All, Advertising) and a window displays with the top factors determining the score for this account.
Are Pipeline Predict and Qualification Scores dynamic? How frequently do the models change / pick up account activity?
Both Pipeline Predict and Qualification Score need to be trained one time. Once trained, Pipeline Predict is updated for all accounts nightly as new activities appear for each account. The Pipeline Predict Score can also decrease when existing activities become stale. Since Qualification Score is a measure of Ideal Customer Profile (ICP) fit, it does not change over time for each account.
How often does Demandbase recommend that I retrain the Pipeline Predict/Qualification Score Model?
Revisiting/retraining the model every 6 weeks is a recommended timeframe.