Design for behaviour change with Microsoft Workplace Analytics
Case Study: Building an early-stage product using Microsoft’s data platform that surfaces insights to drive behaviour change.
Microsoft Workplace Analytics is a product that uses collaboration data from Microsoft Office 365(calendar, emails, time spent on online and offline collaboration, focus time, your top collaborators, etc). This data is combined with organisational context (domains, employee functions, employee hierarchy and org structure) to derive powerful insights about employee behaviour and org culture.
Leaders can use these insights to carry out data-driven cultural transformation in their teams or organisation. Data helps shift organisational culture from something intangible to factual and measurable.
Case study
This case study is for a specific portion of the product, the browsable Insights Dashboard. Instead of just limiting scope to the UI, the design and product team together took charge of exploratory research, data sifting, meaning and insight creation, and ultimately, the final user experience (serving up these insights according our target persona!). The design team was distributed across the globe, so the workshop involved travel and cross-collaboration!
Exploratory Research
We conducted a workshop with cross-functional stakeholders to identify key personas, the scenarios in which they operate, and the data metrics associated with these scenarios.
Key user personas using this tool are CXOs, leaders and senior managers in various organisations. These personas have a few traits in common: They are busy individuals and spend a lot of their day in collaboration with their teams and cross-functional partners. They do not have time for in-depth data analysis, and prefer an overview of critical indicators of their organisation health. Often, they have Analysts working for them, who help them deep-dive into specific areas that catch their interest. They may also have Program Managers who aid them in conducting online and offline change-management initiatives in the organisation.
Besides this, we also interviewed our pilot customers and took the aid of a research study to find what customers really want to know about their organisations.
4 Steps identified to create meaningful insights
Let’s use the example of one of the three focus areas of the Insights dashboard: Employee Engagement. This dashboard would be used by leaders to identify if their organisation has a positive employee experience. Research shows that higher employee engagement leads to better collaboration, more well-being, long-term retention, and better customer experience.
1. What are the critical behaviours?
Few indicative behaviours of engaged employees:
- Have regular 1:1 time with their managers
- Build broad networks beyond core team
- Have strong ties within their team and work in small, well-functioning groups
2. What are the metrics to identify these behaviours?
Metrics for the above behaviours could include:
Management quality and time investment — can be derived from 1:1 time with manager on calendar, presence of managers in other meetings with the employee.
Influence from colleagues — can be derived percentage of engaged employees on the team and collaboration time with them.
Strong-tie relationships — Frequent and relatively intimate interactions with other employees
Work schedule — ordered calendars which are less fragmented, meetings initiated, discretionary effort outside normal work hours
3. Do we have data to build these metrics?
The design, data-science, product and engineering team then got together to identify the data needed to build these metrics, and feasibility. Available collaboration data from the M365 platform was used.
4. Is the data valid? Does it actually reflect known behaviours?
Using a sample set in our own organisation, where Workplace Analytics was running in pilot state, the engineers conducted data validation to identify whether the data was indeed reflecting current behaviours. (Backward testing using a sample organisational data.)
Based on research and an understanding of personas, here is the design principles that we used for the self-guided Insights dashboard:
4 Principles used to design the optimum user experience
1. Design for trust
For people-analytics platforms to gain acceptance and widespread usage in the industry, it is important for users must trust them.
This behavioural data is aggregated at an org level, and employee privacy is of utmost importance. When a leader views these data and insights to see how her company is doing, she should not be able to trace it back to a particular employee. The design needed to bring out this privacy aspect as well, in order to be trusted among potential customers.Here are some key measures to ensure trust:
- Transparency is important — cite data sources, and help users understand how the insight was defined and derived. Always provide on-demand breakdowns and drill-ins and explanations.
- Respect user privacy — Emphasise on data aggregation so that data cannot be traced to the individual. Limit minimum group size for calculation, and emphasise these privacy measures on the UI.
- Show credible insights — Include supporting evidence from industry-standard sources.
- Incomplete is okay, inaccurate is not — During stage-3 of the insights-creation step, we often found that we did not have enough data available at hand to provide a well-rounded understanding of a behaviour. As an evolving product, that is okay. We could always choose to assume and project a more complete picture of the behaviour, but we did not- it was more important to be accurate!
- Suggestions are great, assumptions are not — In order to make the insights actionable, often, it was required to make some suggestions to the user. For example, assuming that employee engagement is low because there are fewer manager 1:1s and suggesting responses to the problem, could make for a very attractive product proposition. However, it was important to take the more humble approach, and present facts as they were — instead of assuming we had the complete picture of the behaviour. This translated into many design decisions: Careful choice of words, careful usage of colours to show data-graphics, and deliberately neutral recommendations.
2. Design for engagement
- Facilitate curiosity- In order for leaders and CXOs to adopt the product, we needed to make sure that the insights were presented and worded in a way which would spark interest.
- Surface few, but relevant insights — The insights needed to be current, and reflect the latest advancements in the people management space.
- Personalise insights — each organisation is different. Personalised insights for various domains (eg: healthcare, finance) and functions (eg: product, sales) help make the product more relevant and engaging for users.
3. Tell a story with data
The insights dashboard needed to be a guided experience since CXOs were busy individuals, and might not have time and expertise to do a deep dashboard drill-in exercise. The data presentation and dashboard design, at its core, was a storytelling exercise. The following questions were asked:
- What is the change we want to drive?
- What is the story we want to tell?
- What are the data pivots to tell this story?
- What is the level of customisation and drill-in to be offered for the particular insight?
- What would be the leading indicators that would help the users do a root cause analysis? ‘good’ graphs and pivots?
- What are the data visualisations and graphs to be used?
- What is the best way to phrase the questions / copy?
- How much explanation is needed?
Based on this, after multiple iterations, a format for each ‘insight card’ was finalised: Key question, data graphics, Insight, Why it matters?, links to supporting evidence, affected locations/organisations/departments, and finally, recommended programs that could be launched for organisational transformation.
4. Design for Action
- Content precedes design — Going back to the meaningful insights creation, it was important for the insights to be crisp and actionable, and for design and user research to play an active role in crafting these.
- Provide a direction over choice — If possible, adopt a clear point of view on what’s the best course of action, and let the UI reflect it. Providing extremely neutral and unbiased representations of the data might be great for trust and transparency, but is not suitable for the key persona, who might be looking for a more guided experience.
Post this, we moved on to the task of UI and content design, which resulted in a finished product, ready to be tested with CXOs and leaders in various organisations.
Working on this project required me to address different spheres of problem-solving.
-How do you distill data such that it makes sense to a particular user and a particular domain?
-How do you design with data?
-How do you create a guided experience for dashboards?
-How do you tell a story with data?
-How do you pivot data such that it leads to behavioural change?
The product that is actually used to create programs for a subset of employees is called My Analytics, which was a different project with another user persona, information workers (or as addressed in the context of this project, ‘employees’)
With cross-team collaboration and using the principles of trust and storytelling, we were able to launch a sensitively designed product which was well-accepted in the industry.
I’ll be happy to answer any questions or comments left below :)
Workplace Analytics is a product owned by Microsoft. The process explained and the views expressed, are my own.
