Focus on the user experience
Success with embedded analytics requires a focus on the user experience, understanding the value analytics brings to each persona, and matching capabilities to users’ needs.
User experience is about more than just pretty visualizations, ease of use, or reducing the number of clicks to accomplish a task. Yes, these things are important, but you can also accomplish even more. Here are five steps to create a great embedded analytics application that will drive both user satisfaction and adoption.
5 Steps to Create a Great User Experience
1. Create user profiles
- – Understand who your target users are as well as their roles and responsibilities. Keep in mind that these may be new users who do not currently use your application.
2. Determine the value of analytics – For each profile, identify the problems analytics can help address, and then qualify and quantify the value analytics brings. Usually, value can be expressed as increasing efficiency and effectiveness, increasing revenue, reducing costs, or improving customer satisfaction.
3. Identify the best-fit analytic experience – Match users to one or more personas that best describe how they need to work with data:
- Information consumers prefer a defined experience in which they view information that has been prepared for them, interact with dashboards and reports, and personalize individual views of this information.
- Content creators want a managed experience where they query governed data sources, create dashboards and reports, and share what they’ve created with colleagues.
- Data analysts need a self-directed experience where they start with a blank canvas, connect to their own data sources, and discover new insights in a more exploratory manner.
4. Match functionality to user needs – With analytics, there’s certainly a lot of functionality that can overwhelm users; this includes any of the visualizations, interactivity, and data that is displayed. But for many users, and especially for those who are just starting out, you should give them only the functionality and specific data they need to work smarter. Release more functionality and data as adoption grows and new questions arise.
- Tip: Use a capabilities map to match users to the functionality they need (see an example on the next page).
- Tip: Produce mock-ups of the functionality you want to implement.
5. Choose the depth of integration – Think about how analytics will be integrated into the user experience of your application. Just as we have seen how consumer web applications embed analytics deep into the context of the application workflow, this could very well be a way to improve the user experience and create a differentiated product.
Taking these steps will help you develop project requirements and prioritize project phases.
User Experience Example
Let’s say you have a set of healthcare applications. These might be the user stories you construct.
- When treating a patient, a doctor wants to see a historical trend of vital metrics, or wants to compare current metrics to a peer group, all within the context of the patient record. This is a purely defined experience with little interaction. Having access to this information makes the doctor much more efficient and improves her ability to provide superior patient care.
- Every time a hospital administrator logs in, he wants to see a dashboard showing critical inventory levels of equipment and critical aging of assets. This is a report he needs to create on a periodic basis. If he sees that items need to be replenished, he can immediately choose to authorize the purchase of new equipment from the dashboard, enabling efficient delivery of patient care while effectively controlling costs. This effort has led to a 5 percent decrease in yearly wasteful spending.
- An analyst might be researching hospital readmissions and looking for new patterns that inform where readmissions are highest and which phase of the patient care cycle needs to be closely monitored. This analysis must be widely accessible for others to create reports. It can to lead greater compliance with industry standard rates for readmissions, resulting in improved patient care and a 10 percent reduction in costs.
After this exercise, look at a capabilities map showing a wide range of functionality and hone in on the capabilities you need.
Common Features in Embedded Analytics
Common embedded analytics features look a lot like common “business intelligence” capabilities, but with a twist: the functionality is integrated into the user experience of the overall application.
The capabilities embedded in each application vary, so we have also indicated how often we see each feature implemented:
Information Delivery – Improving how data is presented to business users is often the main reason software providers want to take on an embedded analytics project. In addition to satisfying users’ informational needs, the look and feel of these capabilities should align with the style of the embedding application.
Dashboards and Data Visualizations:
- A range of visualizations, such as charts, gauges, heat maps, and geographic maps, enables users to quickly draw conclusions and monitor key performance indicators. These can be presented in the context of a single chart or in a collection of visualizations in a dashboard.
Reports: A tabular display of data, often with numerical figures and grouped within categories. Interactivity can include dropdowns and filters for users to view specific slices of data.
Mobile: Capabilities are made available to users on mobile devices, ensuring accurate visual display of information as well as compatibility with mobile device features such as touch input.
Scheduling and Exports: Dashboards and reports can be scheduled for delivery, used in conjunction with thresholds/alerts, or exported to other formats for printing or offline access.
Interactivity – Embedding analytic capabilities inside applications presents interesting ways for users to interact with those capabilities, as well as paves the way for a more informed and productive experience inside the application.
- This enables the user to click on a visualization or report in order to navigate to a different analytic screen or even another part of the main application, and vice versa. In other instances, the interaction simply changes part of the screen rather than the entire screen.
Personalization: Users choose the visualizations or reports which are most important to them, and place them at the top of a dashboard or create bookmarks that can be accessed quickly.
Dashboard and Report Authoring: Users can create their own data visualizations, construct dashboards and reports, and share what they’ve created with their colleagues.
Workflow, Write-backs, and Processes: Analytics is sometimes very tightly integrated with application functionality. Examples include charts embedded on an existing application page to guide user behavior; a report with editable data cells where users can update the displayed data; and a visualization with selectable regions (on a map or area of a scatter plot) allowing the user to perform an action on the selected records.
Analysis – Application providers can enhance the value of the product by giving sophisticated users ways to perform their own analysis, create benchmarks and apply proprietary analytics on their own data, and find innovative ways to incorporate external data sets.
- Users are given data from which they can discover new insights, create custom calculations and metrics, and create new data visualizations. The data set for each end user can be restricted by their user role.
Benchmarking: Users can compare themselves against industry benchmarks and identify areas for improvement. Some cloud applications can even create value for themselves by providing new benchmarks based on data collected across their customer base.
Advanced Analytics: Some applications provide a unique value proposition by developing advanced (and often proprietary) statistical models and making advanced analytics easily accessible in the users’ analysis.
External Data: Some applications incorporate data from external sources and deliver them into a single view or dashboard, turning the application into a vital hub of information. This could be in the form of third-party industry benchmarks, data feeds (such as weather and social media), and customer data from their specific data stores.
Deeper Embedding Leads to Greater Benefits
Embedded analytics strives to bring together insight and action into the same context by integrating analytics deeper and deeper within business applications and workflows. Analytics is embedded within applications in one or more of the ways shown below.
There’s a correlation between deeper integration and greater realization of the strategic benefits of embedded analytics. As you see below, 83 percent of commercial software and SaaS providers who infuse analytics into their applications say they charge more for the embedded analytics in their applications. On the other end, only 74 percent of those who offer a standalone analytics offering charge more.
Therefore, application providers should look to embed analytics deeper into their applications. However, the appropriate level of integration can depend on a number of factors, including your specific use case and resources available for integration. This is why we often find application owners who begin at what they believe to be the easiest point of integration to accomplish in the near term, and then move to the right over time.
In the next few pages, we take a closer look at each stage of the maturity model, when it is most appropriate to implement each stage, and the integration considerations of each stage.
Standalone Analytics Application/Data Integration
The Standalone Analytics Application is Level 0, where the analytics are not embedded into the core application at all.
Standalone Analytics Application
Analytics in a separate application from the process application
Analytics is delivered to users in a completely separate application. This is very similar to the traditional business intelligence model. The only integration is data integration between the main data-generating application and the analytic application. Data access is typically provisioned through a data extract, an API, or a data export.
From the user standpoint, this is a disjointed experience. Users have to work with two separate applications, which likely look and operate differently and have no security integration. A familiar example is exporting data from the application for analysis in Excel, and creating a new copy of data along the way. Once the data changes, the Excel data becomes outdated.
When to Pursue This Model
Common instances of Standalone Analytics Applications include when the product generating data has no business user interface or when the data comes from applications that you cannot embed into. An example of the former is Google Analytics. The service tracks visitor activity on a website. This visitor activity gets fed into the Google data store, but business users can only access that data by logging into the Google Analytics website.
Data Integration Tips
Data integration is essential to any embedded analytics project. Think about these three topics along your embedded analytics journey.
- – Ensure that the analytics solution works well with your data sources today, and be aware of future requirements as your data architecture might evolve over time.
Data connectivity versus data replication – In many use cases, it’s appropriate for the analytics solution to connect directly in real-time to data sources, either via generic connectors (such as ODBC/JDBC), Web Services, or proprietary connectors. In other instances, data should be cached to deliver high-performance analysis without affecting transactional systems.
External data – Increasingly, applications incorporate data from external sources and deliver them into a single view or dashboard, such that the application becomes a vital hub of information. This external data could be in the form of third-party industry benchmarks, data feeds (such as weather and social media), and customer data from their specific data stores.
This is Level 1 of embedding. With single sign-on integration, the main application serves as the user’s “gateway” to the analytics application.
Embedded access- single sign-on from process application to analytics
In this model, the analytics application has integrated security with the core application. Users only need one set of login credentials, which are passed from the core application to the analytics application via single sign-on (SSO). Note that there are still two applications, but the access to analytics is embedded in the core application. It’s still a separate experience, however, because users still have to switch back to the core application if they actually want to put insights to work.
When to Pursue This Model
There are a variety of instances when this model of embedding is appropriate. One example is when you have multiple applications and are creating a single analytic application that accesses data from one or more of these. With single sign-on implemented, users coming from any of the applications can access analytics. This can work well if the analytics application is in the cloud and serving up data from on-premise and cloud applications.
A second example is when you are offering the analytic application as a distinct commercial offering that customers need to purchase separately from your core application, and the “not-fully-integrated” application aligns with your commercial approach.
A third reason is that this model is simply an intermediate step in your development process and you intend to integrate deeper in the future.
Security Integration Tips
Security integration is essential to most embedded analytics projects. Create a secure application with a seamless user experience by thinking through these three aspects to security integration.
- – Ensure the analytics solution works well with your application for single sign-on and passing of roles and rights.
Access control (roles and rights) – It is vital that the analytics solution provides the flexibility and fine-grained control you need to ensure users see only the information and data allowed. In addition, understand whether the analytics solution requires the overhead of continually synchronizing user profiles with the main application.
Multi-tenancy – Understand how the analytics solution provides security with multi-tenancy environments, whether you have separate data stores for each customer, or if data for multiple customers is co-mingled in the same data store (or both!).
Inline Analytics/User Interface Integration
Embedding analytics into the user interface of the application is Level 2. Inline analytics is the most popular form of embedding.
Analytics appear inside the process application (e.g.“reports module”)
In this model, the analytics functionality appears inside the overall UI of the application. Inline analytics is often implemented as a reports tab or module in the application. Another example would be a dashboard on the homepage of the application that users see directly upon logging in.
With analytics integrated at the presentation tier of the application, it is ideal for the look and feel of the analytics functionality to match the UI of the main application.
When to Pursue This Model
Application providers choose this model when users demand easy and frequent access to analytics. Most third-party analytics applications can be embedded using this approach, so there are a lot of options available to “bolt on” to an application. Users are also very comfortable with the reports module approach (as used by Salesforce.com, for example), so it’s not surprising that inline analytics is the most common model for embedded analytics.
UI integration is essential to most embedded analytics projects. Think about these three areas to create a great user experience.
- – Ensure that the look and feel is customizable and can be aligned with your brand. The goal is to provide a cohesive user experience in your application.
Embedding API – Understand the application programming interface (API) utilized to embed analytic content into your application. It should be easy for your development team to implement and maintain over time.
Linking and controls – Often, it is important for users to click on a visualization and go to a specific record in the main application for more detail. In addition, it may be useful when designing the user interface to know how your application controls (such as dropdowns and the like) can be used as input to change analytic content.
Infused Analytics/Workflow Integration
Level 3 means infusing analytics as a natural part of the application. This form of embedding is seeing the most growth.
Analytics embedded within core workflows and application functionality
Here, analytics is embedded within user workflows and becomes a core part of the overall user experience. One way to infuse analytics is to provide analytic content “in the moment” or in existing application screens where users are making decisions and taking action. An example is to provide customer churn risk scores or purchase history in a customer service application so support representatives can offer personalized support.
Another approach is for users to interact with analytic content which leads them to immediately transact or take action based on the analysis they are performing. For a sales manager who conducts geographic territory analysis, this approach enables him to immediately and efficiently re-assign territories from within the analytic interface, instead of having to jump out into the main application.
When to Pursue This Model
This model is for application providers who want to position analytics as a core capability, by bringing together insight and action into the same context. And as we have seen, infused analytics is correlated with greater realization of strategic benefits to the organization.
Workflow integration is essential to the most successful embedded analytics projects. Don’t overlook these three important areas in your project.
- – When analytics is embedded into a page for an existing record, the analytic content should be specific to the record on that page. For example, when logging a support incident for an existing customer, the representative should view analytics only for that specific customer.
Application APIs versus write-backs – If you allow users to transact or update data from the analytic content, you should understand the best way to implement that functionality. Many application providers prefer to call a backend API of the main application that enforces business rules. Other times, you may prefer a direct write-back of data to the backend database.
Security – In regards to transactional and update functionality, it is especially important that only authorized users are allowed to perform such updates.
Level 4 is Genius Analytics, where the user experience is liberating and empowering.
Embedded access- single sign-on from process application to analytics
At this level, the addition of embedded self-service analytics (including data discovery tools) supports unanticipated use cases within a managed, seamless environment.
With embedded self-service capabilities, users can ask new questions of the data as ideas occur to them, leveraging information in ways no software engineer could anticipate. The immediate delivery of insights can accelerate problem solving, spurring innovation and increasing operational effectiveness.
Data analysis becomes integral to not only how people work, but how they think. Efficiency improves and so does decision-making. In essence, companies are creating custom apps for every user—without the heavy lifting that implies.
When to Pursue This Model
This model is for application providers who want to leverage analytics to increase competitive advantage and enhance their applications to deliver unprecedented business value. Only around 15 percent of companies currently offer applications with this level of embedded analytics, making it an expansive field for differentiation.
Embedded Self-Service Integration
When users are empowered with analytics that increase their productivity, they naturally grow to want more. Embedded self-service analytics provides a number of benefits.
Reducing Ad-Hoc Data Requests
When users are empowered with analytics that increase their productivity, they naturally grow to want more. Unfortunately, requests for ad hoc analytics can consume inordinate amounts of your application team’s time, taking focus away from their primary responsibilities. Developers will inevitably need a strategy to meet these perpetually increasing demands.
Deeply embedded self-service analytics are the long-term, sustainable solution. They enable end users to get what they need while the development team stays focused on more important, revenue-generating tasks.
Allowing End Users to Customize Content
Analytics are never one-size-fits-all, making customization a necessity. Embedded self-service analytics, should enable users to customize dashboards and reports without sending requests to IT. This level of embedded analytics also empowers application teams to tailor capabilities to users’ roles and skills, which helps them derive maximum value from applications. Personalizing views ensures every user sees exactly what they need.
Provides a More Complete User Experience
Deeply embedding analytics helps you remove the reasons for users to leave your application. Allowing users to kick off workflows and make changes to the database directly from their reports creates a more valuable experience.
Embedded self-service analytics also enable everyone from your services team, partners, local application owners, and end users to create and share visualizations. Even broadly distributed applications can be tailored to specific companies, teams, or individuals without the burden falling solely on your developers.
Tips for a Successful Embedded Analytics Project
Here are three project management tips for completing a successful implementation.
Have a Vision, But Build in Phases
When you start building specific analytic capabilities into your application, you can become overwhelmed as you see how far you have to go to complete your vision, particularly as new ideas surface along the way. That’s okay. Before you try to take on the whole ocean, remember to start small and build on your successes. Start with one user, one persona, one problem, and one report. Get feedback and move forward. Requirements shift and evolve over time as users start to see what’s possible, so it’s important to stay agile and approach embedded analytics in an iterative way.
Involve Internal and External Stakeholders
There’s nothing more frustrating than building out a really cool feature that no one uses. To avoid this, be sure to get regular feedback from internal and external stakeholders as you build specific analytics capabilities into your application. Utilize screen mockups early in the process and review them with current customers to validate your direction. Ask what they like, what they don’t like, how they would use it, and what suggestions they have to make the product better. This will help you to stay focused on solving real user problems with embedded analytics, and also enables participants to become advocates when the capabilities become generally available.
Perform a Usability Study to Identify Gaps
With select customers, you should consider conducting on-site usability studies to see how they actually use the application and identify remaining gaps. The point of conducting such a study is to find out in advance what problems will bother your users; so be prepared for users to complain when they are lost or frustrated – that’s exactly the kind of feedback you’re looking for. Avoid helping users to get to the right answer. Instead, ask them to complete specific tasks and learn how they expect to navigate your application to accomplish those tasks. Ask them to rate certain aspects of the application and prioritize enhancement requests. Ask open-ended questions to get the most feedback, not questions that can be answered with a simple yes or no.