The old order of traditional business intelligence and analytics is changing, Salesforce brings the cloud, consumer web influence and fast user focused data visualisations that empower the question-answer-question analysis flow.
Analytics Cloud Pioneers
Recently I was lucky to receive 4 days of intensive training, review and product demonstration with the UK Analytics Cloud Pioneers group, led by Anil Dindigal, Director of Cloud Analytics, Salesforce. The Analytics Cloud Pioneers is a group of key specialists from the Salesforce and Business Intelligence partner communities, that have been provided early access to the Analytics Cloud to enable greater insight, product feedback and innovation.
During these sessions the group received a full review of the current capabilities of the platform along with insights into upcoming features and the overall direction of Analytics Cloud. These were long days with a large amount of detailed and complex knowledge transfer and with lots of great hands on activities, distilled down to some key takeaways.
- Analytics Cloud solution represents both a BI platform (the technology that structures and stores the data and provides the analytics engine), and an Analytics App (the visualization GUI that communicates data to executives).
The App and the Platform can be used together or leveraged separately via APIs.
- The App UI is fast, really fast and beautiful. The platform churns over and delivers data at speeds that are beyond any traditional BI expectation.
- Analytics cloud takes a copy of all data for analysis, and supports data from multiple sources; Salesforce data and any other data you chose to throw at it, SAP, ERP, web-logs, etc.
Salesforce data can be scheduled for load though Analytics Cloud.
External data needs some other mechanism to get it loaded, usually this is with an ETL tool, but manual loads and an custom processes vis API are feasible too.
- Analytics cloud architecture is not based on a traditional OLTP relational structure. All data is stored in json data structures as either columnless store or inverted index structures.
- Many of the rules around data have changed, we like to flatten, to denormalise, we like to duplicate data and to transform it, and we like to store data at the lowest grain level; the detail row level. “Within the Analytics Cloud we should treat data like putty.”
- All data values are divided into two main types, dimensions and measures.
- Measures – Numerical values that will be aggregated to generate reportable metrics or KPI’s.
Dimensions – All other data that represents information, most commonly used to group or filter within reports.
- Analytics Cloud platform will handle large volumes of data upto 500 million rows per instance, and reports of millions of rows can run with fast user experience.
- The query engine behind Analytics Cloud (branded SAQL) is built on the technology from the Apache Pig module from the Hadoop project and is based on PigQL, a powerful programming query language designed for analysing very large data sets.
This means we can perform all sorts of complex operations that we have become used to not being able to do in salesforce reports and dashboards, such as dynamic joins, generating temporary datasets, or performing calculations on the fly.
- At this moment of rapid development of the platform, implementations can be highly complex as many of the advanced capabilities can only be achieved by customizing json schemas, developing queries and making changes to the underlying dashboard scripts.
A quick demonstration
The proof is in the demonstration, the thing about Analytics Cloud that anyone new to it needs to experience is the user interface, the speed, and getting the concept of just how much data is being analysed and delivered on the fly.
There are many different use-cases for positioning Analytics Cloud, from full-on global corporate KPI reporting down to data analysis or business executives doing adhoc data exploration. In this brief overview I drill into a data set of the 2014 House Sales from the United Kingdom.
The data set contains 850,000 rows of data, every operation queries against this full data set. Here I am exploring the data from the top down national view and drilling in to identify the grain level details. This exact use case fits well with business scenarios where we are looking to spot trends or outliers.
Here we see the creation of a “lens” a single reporting component, and the the clipping of multiple lenses to create a very simple dashboard, where all items on the dashboard are dynamically bound. All of this from the user perspective can be achieved in just a few minutes.
(*More powerful dashboards do require full specification and technical work to implement.)
Just to set some perspective.
I am not a market analyst and don’t really have the depth of exposure needed to fully communicate what will happen around the BI industry once Analytics Cloud gets some pace, but below are some simple observations:
- The BI industry is quoted as being a $38 billion industry, where-as the CRM industry isquoted as being worth $36.5 billion by 2017. The new Salesforce venture into BI may surpass the scale of its current operations.
- Salesforce are planing a full set of professional certifications (3 + different grades) to support the new platform, providing a new professional infrastructure for a new world of cloud BI.
- Currently Salesforce is the biggest player in the CRM market with a 16% share.
- Currently the biggest players in BI are the traditional on-premise platforms such as SAP 21% and Oracle 14%.
- What will the BI Industry vendors market share look like in 5 years after the Salesforce Analytics Cloud disruption?