What kind of questions can Einstein Analytics Stories answer?

5
(2)

Before getting started with creating first your Einstein Analytics Plus Stories, there are some things you need to think through and decide. Typically people are concerned about data, but we will save that for later. But even before rolling up your sleeves and finding and crunching data, the most important pre-creation task is the need to have an idea of what you want to achieve with your story. In other words, in detail, what is the question you want an answer to? I am sure you have many business questions you want to have answered but if you want to have predictive intel that you deploy to Salesforce as discussed in the previous blog then you need to understand that not every question is suited for the types of predictive models Einstein Discovery uses. So let’s do a deep dive into what questions you can ask the almighty oracle that is Einstein Discovery.

I can ask anything, right?

As mentioned Einstein Discovery can help your business users in their decision processes, so it makes total sense to think about answers that can support those processes. However, you can’t blindly ask any questions and get an answer – as cool as it is, Einstein Analytics is not telepathic and therefore you have to guide and direct all of that analytical power. Questions have to be in the form of minimizing and maximizing an outcome, and it is that outcome that is what the business user will be making decisions. In traditional data scientist parlance, the analysis is detecting patterns in relation to continuous or binary outcome fields. In practical terms, this means picking a variable in your dataset and quantifying it, but in plain language what this boils down to is yes/no questions (binary outcomes variables) or how much and how many (continuous outcome variables).

Bear in mind that Einstein Analytics Stories are not just about building Predictive Models, while those are often the noble end goal of analyses, often Stories are used to answer simple research or confirmatory questions such as “which types of leads are most likely to convert,” “why are they most likely to convert,” “how do these types of leads differ from these other types of leads,” and “what can I do to increase the likelihood that this lead converts” (and so on for every other use case). We’ll discuss the importance of the Story “question” in the overall Use Case, but suffice to say that most reasons for asking these kinds of questions help drive value and efficiency in a business process and not just stop at a few juicy facts.

In reality, you can ask a lot of different business questions and each question would demand a data science project and the building of a model for each. By standardizing the questions you ask we are able to use standardized and proven algorithms and statistical techniques to detect patterns and predictions in your data, but more about this part of the project later. So what does this mean in terms of your questions?

“Yes/no” and “how much”/“how many” questions

When you ask your question you have to center it around a type of observation in the real world but which is accessed via a record you have in Salesforce (or external data, but let’s keep it simple for now) like opportunities, accounts, support cases, etc. In answering the question you have to assess more data than just the question topic, but the question itself should not include multiple variables, hence you can’t maximize or minimize multiple things. Further, you should be able to answer the questions with a yes/no answer or a number. The latter means your question should start with “how much” org “how many”. Looking at some standard questions to ask in relation to Salesforce here are some examples:

Yes/No:

  • Will this lead convert?
  • Will this opportunity close (won)?
  • Will this customer account close?
  • Will this customer churn?
  • Will this invoice be paid on time?

How much:

  • Is the CSAT going to be?
  • Revenue will we make with this customer?

How many:

  • Cases are needed to solve this issue?
  • The number of orders placed in a week?

My questions can be more complex than that

In reality, if you ask your business users what questions would you like to get answered they will give you a longer and more complex question. With a little reframing, you most likely will be able to ask it in a way where you can have Einstein Discovery answer it. Let’s look at an example. Perhaps your business users tell you: “I want to know which of the daily list of cases I need to prioritize first”. First of all, you need to ask what they mean by “prioritize first” unless you already know. And I’d say you should still ask them as your assumptions will seldom be the same as their thinking and we know what assuming does….! So, you dig a bit deeper – are they interested in prioritizing by taking those cases that are likely to be escalated? Or is it some other order of prioritization? Let’s say it really does come down to escalations; those that are likely to be escalated should be first in line to be picked up. You can then reframe the question to “Will this case be escalated?”, which is a question you can easily set up and answer in a story. The story can be deployed to Salesforce and added to your case layout, it will live score any case you view and give you a percentage score Einstein you can then use this score for prioritization; those cases with a high score get picked up first. In plain English, the score will be telling you that some cases are more likely to be escalated than others and here’s Einstein Analytic’s assessment based on the evidence it can see. We will cover how to create your stories and deploy your model later in this blog series. But first, we will look at a methodology to stay on track in your Einstein Discovery project.

How useful was this post?

Click on a star to rate useful the post is!

Written by


2 thoughts on “What kind of questions can Einstein Analytics Stories answer?”

  • 1
    Hayk on February 17, 2020 Reply

    This is the absolute most fundamental article on analytics. One should internalize all of the above to be successful with whichever tool, today or a decade from now! Thanks Rikke.

  • 2
    Vishnu MK on September 25, 2021 Reply

    Your article explained very well on Einstein Analytics Stories usage. Thanks Rikke

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.