Wednesday 11 September 2019

Salesforce Einstein Analytics Capabilities


Einstein Analytics Product (Wave + Prediction Builder + Discovery):

Einstein Analytics in a simplest form is a combination of Wave Analytics, Einstein Prediction Builder and Einstein Discovery

Wave Capability - Pull data from different sources. Only change is; now we can join different tables without using code. Earlier, we used to write SAQL (Salesforce Analytics Query Language) to do inner joins.

Prediction Builder Capability -  It predict the trend as number  (what’s the likelihood of something to happen in percentage) same as Einstein Prediction Builder.  The key difference is; in Einstein Prediction Builder, there’s a limitation that we can’t do cross object predictions, now this limitation is gone because the source of data is Wave engine

Discovery Capability - It tells the narrative (story) of data same as Einstein Discovery, the only difference again is data behind the scene can be enriched as its coming from Wave engine

Einstein Analytics Flow:

·       Data finds the customer  (Data coming from different sources)
·       Salesforce add narrative to that data (Add story around that data)
·       Predict what’s happening  (What happened in the past)
·       Predict what’s likely to happen (What’s going to happen in future, based on trends)
·       Do actions based on insights (Perform actions to change insights in your favour)

Einstein Analytics Assets:

·       App - It’s a collection of dashboards.
·       Dashboard - It’s a collection of lens (reports)
·       Lens - It’s a report, which can use of target dataset
·       Target Dataset - It’s a dataset, which can be used to create lens. It can be created using multiple recipe
·       Recipe - It’s simply a saved set of transformations, or steps, that you want to perform on a specific source dataset
·       Source Dataset - Raw dataset coming from different source for e.g. Oracle ERP, NetSuite etc

Frequency of data refresh:

You can set the frequency of your data to be refreshed on the following basis:

·       Time Based (Minimum is hourly)
·       Action Based (Refresh data based on any action)

Einstein Analytics License Structure

License Type     
Artificial Intelligence
Business Intelligence
Einstein Analytics Plus
1 Billion Rows  
1 Million Rows

Random Points:

·       Minimum data to do analysis should be 5000 records.
·       Not all use cases can be achievable using Einstein Analytics. As of now, it does perditions only in numbers for e.g. (% of likelihood to happen something)
·       As of now, we can’t change the Machine Learning algorithm under the hood, but in future we might be able to plug in our own ML algorithm if we need to.
·       We can launch any action from dashboard which we’ve defined in platform for e.g. trigger, quick actions etc
·       You can define data cleaning rules and can reuse to enrich data periodically
·       Einstein Analytic runs on separate cluster from platform, it copies the data from different sources on its own cluster to store it as big flat file. We can add condition on copying the data, if we don’t want to copy over everything.