The Power of Einstein Next Best Action

10 min readOct 31, 2020

Decision, Decision, Decision…

Review your daily work and ask yourself, how many decisions you have to make everyday? I’m pretty sure we will get to the same result: A LOT! So, wouldn’t it be nice to have someone to help us with that? Maybe directly tell us what is the next best action to take?

Let me introduce you to the concept of “Next Best Action” or “NBA”.

The concept is mostly known in marketing as a customer-centric approach to propose personalized offers or content to the right customer at the right channel at the right time. But you can also think of it as a process of suggesting personalized “best” next steps.

Let’s have a look on a few examples:

Sales — Send custom proposal to a prospect based on likelihood to engage
Service — Send engagement/message to customer with high chance of escalating case to social media
Marketing — Recommend adding a particular customer to a specific campaign

As you can see it’s not only about marketing, we can also use it in all business areas where task automation is applicable.

Einstein Next Best Action

Whenever you need decisioning on what to recommend, when and on which channels…
Einstein NBA offers just that by providing:

Declarative Rules-based recommendation engine where you can build your business rules
Actions to surface the right recommendation at the right time
Possibility to track reactions to the recommendations so that you can make the engine smarter

It’s even getting better, because all orgs receive 5,000 Next Best Action strategy requests per month at no charge. So, you can directly get started! 😊

But before you start, hold back and stay for a few more minutes to let me explain how it works.

Work with Einstein Next Best Action

With Einstein Next Best Action, you can integrate data inside and outside of Salesforce, action strategies, predictive models, your own business rules and insights in one place.

Let’s have a look at the following use case:

During the last sales meeting the sales team told us, that they are fully overloaded with all the different customers. They need our help to inform them, whenever there is a need to follow-up with a customer and what would be the best next step to take for this specific customer. After the meeting we came together with our experts to discuss how we can help the team using the Einstein NBA capabilities. During the discussion we identified three major questions: Who is our customer? What is the best next action for this customer? And, what is the right channel to show a recommendation to our Sales Reps?

Get your data ready

Before we can answer any of these questions, we need to look into the secret sauce of NBA — data. With Einstein NBA we can integrate data inside and outside of Salesforce. By fetching data from external systems, we can get a clear view on our customer across different channels. For example, which products is our customer using and how much money is he spending (Sales), were any cases logged in the past (Service) or what are the different interest groups based on Social Media activities (Marketing). Data integration is a crucial part especially when customers are moving from channel to channel. Once we have all the necessary data available, we can already answer the first question and furthermore, we have a good foundation to answer the second question as well. Because with all the available information we can base our decision about the next best action to take for this customer on data, rather than on our gut feeling.

Create Recommendations

Before we can decide which is the “next best action” for a customer, we need to step by step define what we want to propose. In Salesforce we call this “Recommendations”. Recommendations are an object in Salesforce, and we can create various Recommendation records, that can later be used in our Action Strategy.

Whenever we create a Recommendation, we need to link it to a specified Action. The Action is described as a Flow that is triggered once the Recommendation is accepted.

Together with the Sales team, we discuss what they want to do. We decide to focus on two actions: First, they want to have the ability to offer some discounts to unhappy customer. Second, they want to be informed whenever they should schedule a meeting with a customer.

Let’s create our Recommendations:
Recommendation 1–3: Offer 5%/10%/20% Discount
First, we create a flow combined with a process, that allows us to automatically create the discount offer and trigger a process to send it to the customer. Once this is done, we create the 3 different Recommendations for the different levels of discount and link the Flow as an Action.

Recommendation 4: Schedule a Meeting
Here we also create a flow, but with the option to create a meeting after the Recommendation is accepted. Once done we create a the recommendation and link the Flow as an Action — don’t forget to add a fancy picture 😉.

Create Action Strategies

Great, now we have our Recommendations and we defined the right Actions. The next step is to finally answer our second question: What is the best next action for this customer? Therefore, we need to define a kind of decision engine, that can tell us when to recommend what or what is the best next step. Here we use the so called “Strategy Builder” to build an Action Strategy. To access the Strategy Builder, you just need to go to the “Setup”, search for “Next Best Action” and click on “New Strategy”. Once done, you should see something like this:

To build the Action Strategy inside the Strategy Builder we focus on two aspects: Business Rule and Predictive Analytics.

Business Rules
Before you start exploring the world of Predictive Analytics, let’s first focus on an rules-based approach. Sometimes we can already add value with simple rule-based actions. For example, whenever the Last Interaction with a customer is older than 30 days, we want to show the “Schedule a Meeting” recommendation to our Sales Reps. To create this logic, we just need to load the Recommendation and define the condition using the “Branch Selector” or “Filter” element.

Unfortunately, in the real world, business rules are more complex. Probably, we have a ton of recommendations and we need to decide which is the one to show to the end-user. But for all the rules we want to apply, we can use the different features of the Strategy Builder. For example, we can use the “Limit Re-offers” element to make sure, that the discount recommendation is only shown once a month. Or let’s say you have a lot of different types of recommendation offers that could be surfaced to one customer. Although each type must have its own branch, you only want to surface one type. To do so, filter all your branches through a first non-empty branch and order from top to bottom. Only the first branch that contains recommendations will be shown.

Predictive Analytics
If you are already wondering why we need all the data I mentioned earlier and how we can really come up with a personalized recommendation, rather than a rule-based one, here you go. With the help of predictive analytics (statistics, machine learning…) we can use all this data to create lots of predictions. Likelihood of Escalation, CSAT, Win Rates, Time to Close just to name a few.

When we are talking about building prediction models in Salesforce, we are referring to the following solutions:

Einstein Prediction Builder
With Einstein Prediction Builder, you can make custom predictions about what happens next in your business without writing any code. You can determine what you want to predict, which field represents it, and what object contains that field. This tool is really made for Salesforce admins that helps to easily build predictions using a declarative tool. If you want to know how this works, definitely check this article: “Einstein Prediction Builder: How to Turn Your Idea Into a Prediction” by Thierry Donneau-Golencer, Sr. Director, Einstein Product Management, Salesforce.

Einstein Discovery
If you need to handle a lot of data from external sources and the model needs to be more complex you should go with Einstein Discovery. Einstein Discovery works with Tableau CRM (formerly called Einstein Analytics) dataset which means you can use data from external systems, databases or CSVs. Furthermore Einstein Discovery uses descriptive analytics that tell you what happened in your data, predictive analytics that reveal why it happened (diagnostic insights), what could happen (predicted future outcomes based on statistical probabilities), and what is the difference between variables (comparative insights) and prescriptive analytics that suggest ways in which to improve your predicted outcomes.

Einstein Recommendation Builder (Beta)
Salesforce also recently announced the “Einstein Recommendation Builder” to intelligently recommend records from one Salesforce object to records from a related object. Here are some examples:

Cross-Sell Recommendation: Recommend additional products and services to existing customers.
Next Best Offer: Recommend the most relevant offer or promotion to customers and prospects.
Job Candidate Recommendation: Recommend relevant candidates who are most likely to accept a job, or vice versa.

These recommendations can then be used in the Action Strategy as well.

If we look back to our example, where we want to offer a descent discount to unhappy customers, we need to find out, how unhappy our customers are. Therefore, we can for example predict the Churn Rate of a customer using the Einstein Prediction Builder. The Churn Rate can be influenced by a couple of different information. As this blog post is more focused on the Einstein NBA part, let’s assume we’ve already built this prediction. If you want to know how to do this, please check the resources below. Once the prediction is ready it needs to be available on the customer record in a dedicated field.

After everything is setup correctly, we can meet again with the Sales team to discuss when we want to give the discount. Together with the team we decided to give:

5% Discount to customer with Churn Rate between 30–50%
10% Discount to customer with Churn Rate between 50–70%
20% Discount to customer with Churn Rate between 70–100%

Now we need to add this to our Action Strategy. First, we are loading all the different recommendations for the discount levels. Afterwards we are using the Branch Selector to apply the filter on the Churn Rate and we are creating 3 different branches:

Create branches within the Branch Selector element

In the final step, we are merging everything together, so in the end we can show both recommendations (Schedule a Meeting and Offer a discount) at the same time. The final Action Strategy could be similar to the one below:

Action Strategy within the Strategy Builder

Of course, in the real world your action strategy will be more complex than that one. I just wanted to show you, how you can easily get started though. Whenever you are working with Einstein NBA, you should definitely check the “Generate” and “Enhance” elements. Using the Generate element you can create recommendations on the fly, either from a Salesforce object (an Account, for example) or an external source, such as a SQL database or a company product catalog. With the Enhance element can modify a recommendation on the fly. For example, if a given product is no longer available, recommendations are automatically updated. For more details on how to work with this, check this Trailhead.

Display Recommendations

After creating the strategy in Strategy Builder, we are close to the finish line, but there is one last question: “What is the right channel to show this to our Sales Reps?” Obviously, we want the team to directly get these recommendations, whenever they are opening a customer record and maybe we want to also have the recommendations on an app’s Home page. Luckily with Einstein NBA we have a couple of options. We can use a Lightning record page, an app’s Home page, a Community page, a Visualforce page, or another app, depending on where you want recommendations to appear.

For the “Schedule a Meeting” recommendation, the final implementation could be look similar to this on the Home page:

Source: https://help.salesforce.com/articleView?id=nba_admin_using_strategies.htm&type=5

Here our sales reps can directly see, which customers need their attention. Start with simple strategies and then enhance them over time. For example as a next steps we want to also show the “Schedule a Meeting” recommendation, whenever we are predicting a low customer satisfaction score based on sentiment analysis and historic interactions.

Report On and Track a Recommendation

Before we can hand over the final solution to the Sales team, we should create a custom report type to report on and track recommendation data and strategy metrics. You can see the monthly total recommendations that a Salesforce org’s strategies served. And you can analyze which recommendations are accepted and rejected, who responds to them, and more. This can be used for example, to run A/B tests on two different strategies.

We can also create dashboards like the one below to really keep track of all the important metrics.

Source: https://help.salesforce.com/articleView?id=nba_report.htm&type=5

Congratulations you did the first step towards Sales Automation and now it’s time to present your solution to the sales team. After presenting to the team you are already receiving great feedback and sales is more than happy. They want to already start discussing the next steps with you — so, stay tuned… 😉

Thank you for staying with me till the end! 😊 Feel free to put your feedback in the comments and let me know how your NBA journey goes.

Please find below couple of resources that might be interesting:

For additional help on Einstein Next Best Action, check out Salesforce documentation and the modules on Trailhead

If you have any questions or just want to have a chat reach out to me on LinkedIn.