Earlier this year, we conducted market research on organizational experience, priorities, and cultures related to data and analytics (D&A), and found that data-driven organizations—those poised to capitalize on the promises of analytics—infuse trust for business users at every point in the data analytics value chain, from data collection to data-informed decision-making in solving for real-world business challenges. Diving deeper, this blog series has explored the implications of trust in organizing for analytics, breaking down obstacles to collaboration, and enacting cultural change to make analytics effective. In our final installment, we’ll focus on the details of transmitting trust via the individual experiences users have with analytics.
From our work in customer experience, we believe these four dimensions are required for success:
- Empathy: Experiences are based on an in-depth understanding of the customer—in this case, the business user consuming analytic insights—behaviors, feelings, and motivations.
- Relevance: Solutions and services add value through utility and by meeting basic customer needs – all at the right time and the right place.
- Ease: It is simple for customers to derive value from experiences. Experiences are apparent, accessible, effortless, and uncomplicated.
- Orchestration: Specific interactions and touchpoints are designed and delivered as an end-to-end experience versus a discrete transaction.
Here, we explore ways to think about each dimension in context of the data analytics value chain, and how to use this framework to improve experience and thus trust.
Accessing data to obtain basic information is often an arduous and confusing experience, the first step to undermining any trust that may have otherwise taken root. It is critical to begin from the viewpoint of empathy, which in the access stage is reflected in data provenance, lineage, and definitions that support the interpretation, validation, and use of the data by its human consumers. Additionally, processes and technology to gather and access data must make data and metadata available in a timely manner in order to be relevant, accessible, and valid in the eyes of its users. One practical way to take action here is by establishing a Service-Level Agreement (SLA) for your data warehouse. While this may seem like unnecessary overhead, Scott Breitenother of Casper writes that “an SLA is a promise to your stakeholders that you will deliver quality, predictable service that they can rely on… In short, you need an SLA to build trust.”
Trust is further supported by self-service tools and portals that make access easy, straightforward, and integrated into existing analytics processes and technology. Finally, orchestrating these elements is very important, and considerations of data quality, security, and privacy should be integrated and unobtrusive to the user.
This advice is not revolutionary – of course it’s true that better data quality, timeliness, and self-service are desirable. However, once viewed as elements of experience and key contributors to building the trust on which effective analytics depend, these elements should not be considered as merely “nice-to-have.” If partners served by analytics regularly raise concerns about trustworthiness of underlying data, a quick analysis using the lens of experience may surface areas for improvement.
In the analytics value chain, insight refers to any process that transforms information into knowledge. This can be as simple as displaying a sorted table from a database, or as complex as the output of a sophisticated predictive model using deep learning.
The work of delivering trustworthy insight must begin with the end in mind, requiring a deep understanding of the user’s business problems and the available levers she has to take action on the insights delivered. This puts empathy in the driver’s seat: understanding why the user is consuming the information, how they want to consume it, and where they consume it from must inform everything from the initial lines of analytical inquiry through to the final product. Particularly at the point of consuming information, analytical insights are experienced as relevant when they are provided in context of work rather than locked up in a separate reporting system. For example, suppose an employee on the Finance team is responsible for approving expenditures based on an analysis of historical spend and budgets. In the decision process, he or she may need to consult several sources of data including emails, past invoices, and budget trackers. In a seamless experience, the employee would have access to all of the reporting needed to make an approval decision in a single location. This frictionless experience is grounded in an intimate understanding of user behaviors, processes (both documented and undocumented), needs, and jobs to be done to translate information into action.
Care for the user’s experience must also pervade design choices at every step of data analysis. In the context of building trust in analytics, the dimension of ease can be represented by favoring simple, explainable, and intuitive analytical models. Advanced models will certainly mature past basic decision trees and linear regressions, but this maturation must proceed at the same pace as the consumer’s willingness to trust and act on the model’s outputs. Orchestration of the end-to-end experience is then achieved by providing selective transparency into the inner workings of the model, for instance by arranging for interactive testing and tweaking of model parameters when the model is being presented for validation and acceptance (and thus considering this factor when selecting the tools and formats used for communication). This extended and interactive engagement with end users is also the ideal opportunity to practice empathy in the form of data storytelling and discussing the business meaning of interesting analytical results; artfully exposing carefully selected and tractable elements of an analysis along with the final outcomes is crucial to building trust in analytics that may otherwise be opaque.
Knowledge is power, but only once it is acted upon; the analytics value chain extends through the action that is taken as a result of insights being delivered. If any of their work is to be worthwhile, analytics functions must take more responsibility for the experience of how insights are translated into action, ensuring that solutions do make a difference in business operations.
Starting again with relevance and empathy, analytics teams must start with understanding roles, workflows, and jobs to be done: in short, what information users need and the corresponding levers they can pull to manage their business. Relevance at the point of taking action means delivering insights which empower users to add value within their business role – that is, improving their ability to uniquely contribute value to business success. This may sound basic (and it is), but in North Highland market research conducted in January 2018, 47 percent of respondents still report that requirements for analytics solutions “center on how to deliver existing data, analysis, or insights” more often than that they “originate from business measurements of success,” so there is still much opportunity to become better attuned to empowering internal customers. Hand-in-hand with understanding user roles is understanding users as people, an empathetic understanding of their motivations, incentives, and other barriers to action when they have actionable information in hand. Will a user suffer negative personal consequences if his decision meets with poor results, even though it was justified by the available insights? Would a user feel confident justifying his or her insights-based decision to management? These questions extend beyond the purview of analytics, but are critical to understand if insights are to be put into action.
Equipped with this understanding, solutions are clearly most effective which are easiest to use – where technical frictions in translating insight into action have been reduced or removed altogether. There is room for improvement here as a similar 47 percent of respondents to the above cited survey reported that “analytics insights are pulled via report or dashboard outside of the normal flow of work” more commonly than that “analytics insights are delivered within the context of actions or business decisions.” Finally, the orchestration of this experience extends beyond integration of one-way insights and takes the form of integrated feedback and continuous improvement. This is particularly important for modern machine learning applications which feed on data, since feedback allows teams to not only measure but also to rapidly improve the effectiveness of solutions. This continuing engagement with production systems that are measurably delivering value creates a virtuous cycle that empowers both the analytics team and end users.
The process of experience design helps ensure that all considerations for a product or service are captured – and at a time where companies are increasingly looking to use data and insights as a source of competitive advantage, using an experience design lens is justified for ensuring no source of value is left untapped in the design of analytics solutions. Since experience is also a key component of trust, experience design—shaped around the core principles of empathy, relevance, ease, and orchestration—plays a critical role in achieving repeatable and real business value from analytics.
For a complete picture of the landscape of trust in analytics, take a look at the other pieces in our series: