Posted by Eric
May 28, 2020
As Business Intelligence (BI) has become increasingly integral to business operations, embedded analytics is a key area of focus. In this blog I delve into the advantages of embedded analytics, the multidisciplinary complexities involved, the types of embedding environments, and the degree of integration. Most importantly, I share essential questions that organizations can ask upfront to facilitate high-quality, timely, dependable analytics capabilities within their operational infrastructure. These questions are based on my considerable experience gained on many consulting engagements that involved embedding Tableau into end-user products to enhance the analytics experience.
Embedded analytics involves the integration of a third-party analytics tool within another application to provide interactive data visualization and advanced analytics capabilities such as machine learning, statistical modeling, etc. With embedded analytics the analytics tool becomes part of the normal user workflow within the hosting application. The main differentiator for embedded analytics is that the user does not “switch” to another application… they stay within the hosting application.
Because of the server-to-server or service-to-service nature (or any combination of the two) of embedded analytics, there are inherent complexities that span multiple disciplines within an IT organization. For example, successfully embedding analytics requires network, hardware and data architecture, authentication integration, authorization modeling, data visualization, business analysis, data integration and preparation (i.e., ETL), workflow/process automation, quality assurance, application operations/support, etc. In addition, complex concepts frequently become part of the conversation such as multi-tenancy, privacy regulatory compliance, multi-network connectivity, and security, etc.
While most semi-technical or non-technical staff can use and be productive with data visualization tools, embedding data visualizations within another application requires “hard core” IT skills. Due to the complexity of the various scenarios and the large number of “integration points” between IT systems, embedded analytics is frequently a substantial undertaking for IT organizations.
There are a few common scenarios when it comes to embedding analytics. This section is not a comprehensive list of all possible embedded analytics scenarios but does describe some of the more common high-level situations.
One of the most common differentiating scenarios with embedded analytics is what the hosting application is. The hosting application will likely be either a custom-built application or a third-party application.
With a custom application, the organization desiring embedded analytics seeks to use the analytical capabilities within an application they are creating themselves. Custom applications are coded or assembled by the embedding organization’s internal resources and the third-party analytics tool is visually integrated with the custom application.
With custom application embedding there are complex considerations such as:
Embedding within a third-party application is another common scenario. In this situation, the organization is often embedding a third-party service within another third-party service. For example, embedding Tableau Online visuals within Salesforce, or embedding Microsoft Power BI visuals within Microsoft SharePoint Online.
When embedding within third-party applications there are somewhat different considerations that must be evaluated, such as:
Another key consideration is how tightly integrated the embedded analytics will be. Does the analytics tool appear as a fully integrated, visually seamless component within the environment? How much interaction is required between the hosting application and the embedded analytics tool?
In some scenarios the analytics tool can just be “plopped” into a webpage with minimal effort. While in others the user should have no indication whatsoever which parts of the user interface come from the hosting application versus the analytics tool.
To achieve seamless integration often requires a lot of “nitpicky” design and integration. How “skinnable” is the analytics tool? Can you achieve the same colour palette as the hosting application? Can buttons or controls have the same visual appearance as the hosting application? How can data be passed in both directions (i.e., into the embedded visual and out of the visual into the hosting application) without additional steps on the user’s part?
These types of questions need to be considered before even selecting the analytics tool as different tools have differing degrees of these capabilities.
In some situations, full, seamless embedding may not be necessary, and the data visualization tool can just be dropped into the hosting application. No bi-directional API integration is required, and only “out of the box” user interactivity provided by the analytics tool is needed. In this situation, embedding can be accomplished with very little effort in a matter of minutes.
A critical precept to achieving optimal results with embedded analytics is to “walk before you run”. What is the minimum amount of work to achieve one of the simplest deliverables? If you try to wait until all possible deliverables and scenarios are defined and accounted for, you may never achieve a deliverable or require a timeline that is just impossible for the organization to accept. This approach is usually referred to as the minimum viable product (MVP), which is a technique for quickly learning lessons needed to achieve the minimal deliverables that early adopters of the capability might require. The MVP approach allows for more agile integration of user feedback when designing the analytics capability and of overcoming increasingly more complex embedding concepts over time when seamless embedding is required.
Once an MVP deliverable has been defined, there’s a general process that needs to occur to achieve that MVP and loosely follows this sequence:
Embedded analytics is a significant focus area for most organizations. Prior to 2008 most analytics tools were standalone and required switching between business applications and the analytics tool. As business intelligence (BI) has become a more critical aspect to organizational operations, tighter integration is demanded. While embedding is a more technical undertaking, following a well-defined process and asking the right questions up-front can help organizations achieve high-quality, timely, dependable analytics capabilities within their operational infrastructure.
You can embed and customize Tableau into your offering to create a seamless experience for your customers LEARN MORE
If you need guidance embedding Tableau in an end-user product such as Salesforce, web apps, portals, hosted or on-premise software products, CONTACT US
Explore Posts By Category
Recent PostsThree Table Calculation Functions for Tableau that Should be on Your Radar What to Ask Before Embedding Analytics with Tableau 7 Most Impactful Business Intelligence Trends for 2020 More than a Business Intelligence Maturity Model Tableau Master Challenge #6 - Reference Lines and the Order of Operations
Need help embedding Tableau within another application? We can assist you - leave your email!