Driving sales efficiency with Dynamics 365 and Microsoft AI

Dynamics 365 for Sales and Microsoft AI are used to present the highest-priority sales leads to sellers. With this new solution, sellers can quickly act on the leads most likely to become customers. The solution creates a simple, intelligent system for discovering and contacting the best lead and makes our sellers more efficient and effective.

In today’s world of selling, Dynamics 365 and AI form a powerful combination that helps sales professionals stay as productive as possible. Microsoft Digital used Dynamics 365 for Sales and Microsoft AI to create a solution to present the highest-priority sales leads in the system to sellers. With this new solution, sellers can quickly act on the leads most likely to become customers. The solution creates an efficient and simple system for discovering and contacting the best lead.

Finding high-quality sales leads

To generate global demand for Microsoft products and services, the marketing and sales organization collects leads when people request information or access gated content via an online form. The system collects leads by using marketing vehicles, such as:

• Trial products, services, and subscriptions for products like Azure and Microsoft Office.
• Events like conferences, webinars, and training sessions.
• Content downloads.

When someone signs up for a Microsoft product trial, sends us an email, or downloads content, the person or business becomes a lead for the purposes of sales and marketing. We market to over 10 million leads per year. These names—which might be company or people names—become records in the marketing and sales system. The salesperson takes the lead received from the marketing and sales system and works to convert the prospective customer into one who’s ready to buy Microsoft products or services.

Building lead qualification

As a sales organization, Microsoft has the primary goal of being as effective as possible in taking leads and converting them to customers—that is, to users and purchasers of our products and services. The existing lead-qualification solution began the process by acquiring the lead information and ensuring that each lead was a potential customer with valid contact information. The solution then determined the intent of the lead. It profiled and nurtured the lead in preparation for a marketing qualification process that used AI and an email bot to determine the lead’s intent, move the lead to an identified sales opportunity, and finally, move the lead to a sales win and to customer success. Find out more about this solution in the paper, Microsoft increases sales by using AI for lead qualification.

This marketing process provided a better assessment of our leads, but it still left a significant gap between the handoff from the marketing solution and the point of engagement by our inside sales team. We needed to further determine the eligibility of a lead for potential sales and route the best leads to our sellers as efficiently as possible. The following figure depicts the process of building lead management for sellers.

Examining the need for higher-quality leads

Although the preexisting system removed poor-quality leads and provided basic lead qualification, we recognized the need to do more for our sellers to give them the best leads to work with. The sellers were still working with numerous leads that didn’t convert to customer successes, because the leads simply weren’t ready. The sellers struggled with several issues:

• Limited capacity. The sellers have set working hours each day, and their time is valuable. We found that the majority of the sellers’ day was spent following up with leads that didn’t move to successfully converted customers.
• Massive volume. The sellers were presented with almost 10,000 leads each day that met the qualifications to require contact with the lead. We didn’t have enough seller capacity to handle leads in a timely manner, and the huge number of leads still contained many that weren’t high quality or ready to become customers.
• Missing information. The leads presented to the sellers were qualified and sorted, but sellers still had to determine which leads in their lists were the best. Sellers often manually assessed leads without a complete context.

The company wasn’t using the sellers effectively. About 18 percent of the leads converted to customers, and we wanted to increase that rate. The sellers needed access to simple, accurate, and well-qualified leads. We needed a way to send the very best leads to our sellers so they could convert those leads to customers.

Establishing the next best lead

We wanted to create a solution that would take the leads from our marketing system and give our sellers simple, actionable access to the next lead in the system that provided the best opportunity to create a Microsoft customer. We set several goals for the solution:

• Incorporate the existing lead-prioritization solution. The existing lead scoring and qualification solution using a bot had generated significant progress, increasing the lead conversion rate from 4 percent to 18 percent. The idea was for the new solution to build on that progress and become the next component of a larger lead ingestion and processing solution that would generate even higher conversion rates.
• Create a simple and clear action for the sellers. We wanted to eliminate manual filtering and other tedious tasks for the sellers. The goal was to create an interface that simply put the best leads in front of the sellers.
• Surface the most actionable and highest-quality leads to the sellers. The root goal was to put better leads in the hands of the sellers—to improve the lead conversion rate and help ensure that high-quality leads always made it to the top of the millions of leads coming in.

Capturing true sales intelligence

We evaluated several options for implementing the next phase of our lead management solution. We developed plans for three core components that contributed to the new solution: a lead scoring engine based on machine learning to better categorize and catalog incoming leads, a system for managing and applying business rules to qualified leads, and a Prospect Store to group and rank the leads ready to be presented to sellers.

At the core of the solution are several key features we designed to effectively guide the decision making of our sellers:

• Real-time, dynamic lead prioritization that automatically sorts leads based on the likelihood to buy, incremental revenue, and consumption.
• A continually updated system that guides sales decision making with the latest information.
• An intelligent evaluation of the return on investment to determine optimal engagement strategies based on the cost of acquisition.
• AI trained to use available data to make holistic assessments.
• An optimized seller interface that accounts for seller scheduling to provide appropriate points of contact.

The following figure depicts the key features that guide the decision making of our sellers and capture true sales intelligence.

Creating better lead ranking with AI-based lead scoring

We created a lead scoring engine to get more accurate and relevant rankings for the existing leads in our marketing engine. The lead scoring pipeline ingests leads from the marketing engine into an API for the lead scoring solution. The API pushes the leads through machine learning models hosted in Azure and returns them back to the marketing engine with a lead score. The lead scoring solution examines over 200 data points within the context of the lead, including:

• Personal information:
  • Does the lead have a valid phone number?
  • What’s the quality of the phone number entry?
  • Does the lead have a valid email address?
  • What’s the type of email address—work (such as .com) versus educational (such as .edu) versus personal (such as outlook.com)?
  • Is the email address similar to the lead name?
  • Does the lead have a valid contact name?
  • Does the lead align to a company with a valid name?
  • Does the domain come from regional spam filters?
• Firmographics and demographics:
  • What industry does the lead belong to?
  • What’s the size of the company?
  • How senior is the lead (such as manager or director)?
  • Is the lead a partner?
  • What’s the lead’s job title?
  • How senior is the person who created the lead?
  • Does this lead belong to an account that has a designated account manager?
• Product and service usage data:
  • What is the Office trial usage?
  • What is the Azure usage?
  • What is the Bing Predicts data?
• Marketing interactions:
  • How old is the lead?
  • What was our previous interaction? When did we last have an interaction? How frequent were the interactions?
  • What’s the source campaign?
  • Has the lead requested contact?

The model runs several machine learning algorithms to examine these data points and assigns a numeric score to the lead. As the data changes, the model continually reexamines and rescores the leads to provide the most accurate and current results. The solution uses the model score to determine whether the lead is ready to move to the next stage in the lead qualification process. The lead management process sorts the leads by prioritization score and forwards those meeting the scoring threshold to the next level of the lead routing process. Machine learning models exist for different scoring situations, including those for:

• Regions
• Languages
• Products

The following figure depicts the components that comprise AI-based lead scoring process.

Matching leads to sellers via business rules

We also created a business rules management system that uses specific policies to match leads with sellers. Customers—and sellers—speak different languages, and sellers have product expertise they use to help customers make the best product choices. The business rules management system is designed so the leads most applicable to a seller will become available to that seller as potential leads.

Business rule policies can contain numerous rules and conditions so a customer won’t be contacted outside of normal business hours or too frequently. Policies are based on lead and seller data, including:

• Calling hours based on region. This condition determines the time of day during which our sellers call a lead. It typically corresponds to the working hours of the customer and the time zone in which the lead resides.
• Call frequency. We use this condition to limit the total number of calls a lead receives from our sellers.
• The amount of time between calls. This condition limits the frequency with which we contact leads.
• The seller’s specialty. Knowing a seller’s specialty helps us connect leads to sellers who are experts in the products being pursued.
• The seller’s region. This condition helps ensure that our leads are contacted by sellers who are fluent in their language and who can address any regional specifics for a product.

Finding the next best lead via the Prospect Store

The Prospect Store provides the filtering functionality that determines which specific leads will the solution presents to sellers. The Prospect Store ingests leads from our marketing system and creates a first in, first out queue containing leads. The business rules manager runs the leads—which the lead scoring engine has already accurately scored—for each seller and surfaces those leads through Microsoft Sales Experience (MSX), the Dynamics 365 for Sales instance used by our sellers for follow-up. From MSX, the seller requests the next best lead by using a simple interface. The Prospect Store allows sellers to use both new and follow-up leads and creates a lead record for the next best lead in MSX. The Prospect Store contains several components that work together to provide complete functionality:

• The integration API. The integration API interfaces with the internal Prospect Store database and enables insert and update actions for incoming lead data.
• Prospect Services. Prospect Services is built on Azure Service Fabric and provides the core functionality for lead storage and surfacing. Prospect Services includes two components:
  • The Prospect Store database. The Prospect Store database is built on Azure Cosmos DB and stores the key information for leads that the solution manages and surfaces within the Prospect Store.
  • The State service. The State service tracks the state of leads in the Prospect Store to help ensure that only active leads are surfaced in the Prospect Store. Leads can be on hold for various purposes, such as the lead being engaged with another seller for a different Microsoft product or service.
• The Next best lead service. The Next best lead service surfaces the next best lead for the Inside Sales seller.

Using triggers for data management

For each change that occurs within the Prospect Store, we generate events that are used for communication among Prospect Store components. A trigger component creates an event in an Azure Service Bus topic for each change that occurs. Azure Service Bus provides a central point of communication that all components can easily access. External components, such as reporting systems and monitoring systems, can also subscribe to the topic and consume the events various purposes.