Why your organization needs user-centric analytics – and how artificial intelligence will help you build it

Modern businesses increasingly rely on analytics to stay ahead of the competition. More than ever, they need clear insights to achieve their goals. Artificial intelligence has become a powerful ally in this quest for clarity.

But AI is not a singular application. It can be applied separately in every link of the analytical value chain – not only in data science and integration, but in user interfaces as well. In fact, combining AI and UI will lend focus and direction to all other aspects of the value chain. This is the essence of user-centric analytics.

Data-centric design has long been a mainstay of analytics. For generations, we have organized essential information into categories and topics, grouping KPIs and presenting them to users in subject-specific reports and overviews. In many ways, these user interfaces resemble a visit to a traditional library. There are endless stacks of perfectly organized reports for us to browse and it’s up to the users to find the data they need to achieve their goals. Need three separate statistics? Find three separate reports.

This approach made sense in the past, in a time when structuring information was a purely human job. Creating universally intelligible categories and letting the users do the legwork was far more pragmatic and economical than hiring an army of librarians to create personalized overviews of diverse datasets. With the advent of artificial intelligence and machine learning, however, that arithmetic has changed.

Today, user-centric design doesn’t only make more sense than data-centric design. It outperforms it. After all, business goals are achieved by the users – not the information. Giving your users access to intelligent interfaces that proactively support those goals and anticipate their needs will vastly increase their efficiency and make your organization more responsive.

Business goals are achieved by the users – not the information.

Everything starts with understanding your users

At first glance, asking your data scientists to design your dashboards seems like a sensible choice. They often understand the details and intricacies of data better than anybody else. But while such dashboards would certainly be incredibly thorough, it is likely that they would often describe the world as data scientists see it. And the reality of the matter is that many of your users won’t be data scientists.

What your users need most is an interface that is explicitly optimized for their needs.

What your users need most is an interface that is explicitly optimized for their needs. These often vary considerably. There is a wide range of user groups and personas across all levels of your organization. They each interact with data in their own ways, with KPIs and goals that are unique to their specific roles and responsibilities. Understanding these users is the first step toward designing analytics interfaces that truly enable and support them – and doing so requires a specific set of skills. It would be unfair to expect this of your data scientists alone, who are already a rare breed as it is.

Embracing the core tenets of user-centric design will help you determine what your analytics UI should be capable of. What drives each user group’s performance objectives? How do they determine whether they are on target? And which actions do they take to intervene if those targets are not being met? Armed with this knowledge, you can construct analytics interfaces that support the goals of each user persona perfectly. But more importantly, you can combine this knowledge with the powerful capabilities of AI to deliver data science and insights proactively.

<p>Why AI-driven UI is the Future of Analytics - Target Icon - by Accenture</p>

AI doesn’t magically know which KPIs or actions it should recommend to your users, but the basic building blocks of user-centric design work just as well for AI-driven UI as they do for conventional interfaces. Clearly, the value of a UI that recognizes needs before they arise and seamlessly suggests appropriate actions to each user speaks for itself. But building that intelligent interface requires insight into objectives, context and actions. It all starts with understanding your users.

Endlessly adaptive and intrinsically helpful

Imagine the following scenario: a multinational consumer electronics retailer experiences a dip in their European sales. A competitor has launched a new product in Germany and is eating into their local market share. Try to put yourself in the commercial director’s shoes. What type of analytics interface would best suit your needs? All the data is right there in the system. Would you prefer to drag it out yourself and draw your own conclusions? Or would you rather have the system guide you in the right direction, automatically drawing your attention to the issue, illuminating the underlying causes and setting up a call with the regional sales director to discuss options and strategies?

Once you define the responsibilities, goals and actions for each group of users within your organization, you can use artificial intelligence to deliver insights and facilitate decision-making. Accenture has developed the Exception-Drilldown-Action model to illustrate this and help you guide your UI design. Originally conceived for visual user interfaces, this approach has proven equally valid for AI-driven UIs.

Why AI-driven UI is the Future of Analytics - Exception Drilldown Action Model - by Accenture

1. Exception

When your analytics solution discovers an exception, it should trigger a response. What constitutes an exception for any given user depends on their goals and responsibilities, as well as the goals of your organization. For instance, if the system determines that one of your sales representatives is on the verge of losing an important client, it should help them take steps to remedy the situation. Our earlier example of dwindling sales would be an exception for the commercial director.

2. Drilldown

A drilldown clarifies why an exception has occurred. In general, there are two types: why and where.

‘Where’ drilldowns break down the primary KPI (e.g. sales vs. budget) along key dimensions. Usually, this follows organizational structures like product categories and geographies. This helps users identify who among their direct reports is most likely to have actionable insights (e.g. country or product managers). It also allows users to focus on the target group where their efforts will have the greatest impact. Artificial intelligence and machine learning can help identify and define these target groups, especially in cases where the data doesn’t follow the organizational structure (e.g. marketing analytics).

‘Why’ drilldowns examine different metrics that go beyond the primary metric. They seek to explain the drivers that influence the primary KPI (e.g. sales), such as market share, customer satisfaction, marketing effectiveness indicators and other metrics. Again, AI and machine learning have much to offer, as they will help inform the UI which metrics have the greatest impact on the primary KPI.

3. Action

Once an exception has been raised and the causative forces behind it have been identified, your UI should help users formulate an appropriate response. These actions, of course, are as diverse as the exceptions that precipitate them. For a sales representative faced with the prospect of losing an underserved client, the system might suggest rearranging or canceling appointments with other clients to ensure their at-risk client receives the attention they deserve. For a commercial director faced with flagging sales due to increased competition in a specific product group, however, a direct call with the regional product manager may be more appropriate.

These are only a few specific examples, of course, but our experience shows that the Exception-Drilldown-Action design model is universal – provided it is applied with context in mind.

User-centric analytics with visualization and speed - by Accenture

Putting user interactions in context

When it comes to designing user-centric interfaces, visual analytics solutions can seem like a foregone conclusion. After all, visualizations are an excellent way to convey insights and express complex data. But a visual representation may not always be the most appropriate or useful approach. It limits your thinking to only one channel and one mode of interaction. And if you want to avoid this, context is key.

Executives rarely have time to sit at their desks and use their computers. They are often on the move and need their sources of information to move with them. Including voice-based interactivity in your business analytics will allow them to access the insights they need when they need them, without being limited to screen-based environments. Likewise, incorporating a digital assistant into your analytics suite will help your sales representatives manage appointments and review client information on the road, increasing their productivity and efficiency.

On the other hand, employees working in busy office environments will have less use for voice-based features. In these cases, a visual approach is the more appropriate choice. But all these scenarios do have one thing in common: if you want to provide each user group with the ideal user experience, your interface design must be capable of adapting to the context in which the solution will be used.

User-centric Analytics - Computer Icon - by Accenture

Always let user interests guide AI-driven UI

Making user interfaces more intelligent will allow you to deliver better user experiences. You can only do this when you understand the user – and by extension, their goals. Artificial intelligence is an exciting tool in this regard, both for gaining insight into user journeys and crafting solutions that make those journeys more seamless. But it is not a goal in and of itself, nor is it limited to just one application.

From data integration and data science to the user interface itself – AI can fundamentally augment the way your organization handles its key processes. It won’t do so out of the box, of course. There are no universal solutions. You will have to experiment and learn as you go. Try often, fail fast, build new iterations on the foundations of past successes. Seek advice and practical assistance as necessary to overcome the challenges and roadblocks you will encounter along the way.

AI and UI is what will take user experiences to the next level.

But most of all, you should remember that where the impact and value of your analytics projects rely on users taking action, it’s user experience that is the key to achieving success. And in a world where data science will increasingly determine your competitive edge, merging AI and UI is what will take those user experiences to the next level.

Author: Accenture the Netherlands