Skip to Content
Product Information
Author's profile photo Janet Nguyen

Exploring your Data with Smart Discovery

Smart Discovery is a powerful feature within SAP Analytics Cloud that uses machine learning to explore your data and uncover valuables insights.

The feature enables you to:

  • Discover the key influencers driving your KPIs
  • Gain insights about these influencers
  • Identify outliers
  • Analyze patterns in your data
  • Use historical data to predict future outcomes
  • Simulate ‘what-if’ scenarios

Using Smart Discovery

To get the most out of Smart Discovery, you’ll want to have questions in mind such as:

  • What are the key influencers driving deal size?
  • How do different regions impact the number of licenses sold?
  • What correlation, if any, exists between supervisor and contractors’ ratings?

In other words, what aspects of your business do you want to explore through your data?

Looking at a fictitious example, let’s suppose we run a global software company that employs hundreds of people. We have sales managers, full-time employees, and contractors. We sell software licenses to three customer groups:

  • Fortune 500
  • Large enterprise
  • SMBs

We start our Smart Discovery analysis with a question — what are the key influencers driving deal value? We are essentially trying to determine what factors contribute to the number of licenses companies purchase.

Using SAP Analytics Cloud, we login and upload our sales data. We choose to create a new story from the menu and are presented with all the measures and dimensions from that dataset.

We can select a single measure or dimension and start finding insights. For this example, we will select Deal Value from the list, then exclude all the measures and dimensions that we don’t want included in the analysis. Now we can run our Smart Discovery and begin our analysis.

After running Smart Discovery, three main page elements appear:

  • Key Influencers — displays the measures and dimensions contributing to our success
  • Deal Value Size — displays the deal value size based on the key influencer we select
  • Insights — displays key insights based on the measure or dimension we select

By selecting one of the key influencers, we can focus our analysis. In this example, we have selected Customer Segment, which changes the display for the Deal Value Size chart. Now the different customer segments are displayed: Fortune 500 companies, large enterprises, and small to mid-size businesses (SMB). We can see what percentage of our total customer base each of these segments make up, as well as see which ones are above and below the average deal value.

Looking at Unexpected Values

We can dive deeper into Smart Discovery by looking at the unexpected values. This means that the system has predicted future values based on our data, but has missed the mark in some way. These predictions happen automatically and are calculated in Smart Discovery without the need to run a separate predictive forecast.

In the figure above, we can see there are nine instances where the predictive forecasting identified records that were different than expected. In the first row, we see that the expected deal value for one of our vendors was projected to be $85,386.41, but the actual deal value was only $11,000. Naturally, we are interested to find out why this was the case.

Alternatively, if the actual deal was higher than what was predicted, we may similarly be curious to learn about the contributing factors so that we can potentially replicate this success in the future.

Smart Discovery also displays a scatter plot, showing us all our deals. We can select any data point on the graph to learn more about the deal. This is useful if we want to learn more about the outliers identified by the Machine Learning algorithm.

Running a simulation

We can also run a simulation using Smart Discovery, which allows us to test hypothetical scenarios. The chart displays all the measure we selected, and excludes all the ones we didn’t select.

Along the bottom of the chart, we have eight measures:

  • Base Value — defines the expected value size excluding all variables. In this case, our base value is $15,505.38. Any additional variable will either enhance this or detract from this base value.
  • Number of Licenses — represents the expected value additional licenses will bring. All else remaining equal, if there are more licenses, then there will be more revenue added to the base value.
  • Customer Segment — represents the additional value different customer segments bring. Here you can isolate each customer segment to refine your analysis.
  • Number of Customer Meetings — represents the additional value customer meetings bring. Here you can determine how customer meetings contribute to revenue.
  • Length of Sales Cycle — represents the additional value different stages in the sales cycle bring. Here you can determine which is the best sales cycle length.
  • Contract Level — represents the additional value contract levels bring. Here you can determine which contract level contributes most to your revenue.
  • Country — represents the additional value each country brings. Here you can determine which country contributes most to your revenue.
  • Deal Value — represents the total deal value taking into account all selected measures and dimensions.

Suppose we want to know what, if any, correlation exists between revenue and the Number of Customer Meetings with Fortune 500 C-level clients in North America.

Some questions we may want to ask are:

  • What impact do customer meetings have on revenue?
  • Is there a point of dimensioning returns?
  • What is the sweet spot, or optimal number, of customer meetings to have with C-level clients?

We begin by modifying the Number of Customer Meetings by dragging the slider all the way to the left.

Immediately, the accompanying chart adjusts. Having only two customer meetings decreases revenue considerably.

If we drag the slide all the way to the right and increase it to the max, it dramatically increases revenue.

But is this the optimal number of meetings? To determine the so-called sweet spot, we need to play around with the slider a bit more.

Factoring in all other variables, we determined that 40 meetings is the optimal number of meetings to have with your Fortune 500 C-level clients in North America. Any more or any less will detract from our total revenue.

The results can be pasted to new or existing pages, and you can take action by sharing the insights within your organization.

Additional Resources

Assigned tags

      Be the first to leave a comment
      You must be Logged on to comment or reply to a post.