Designing Dashboards Your Users Will Actually Use

How many times have you developed dashboards for your product that your users never actually use?  As product managers, we have to design features for maximum impact.  All too often dashboards are the red herring of product roadmaps.  Everybody wants a dashboard, but nobody uses it.  This can be a huge investment with little return for sales and retention.  So how do you design a dashboard that will be used for more than a sales demo?  As a UX Designer turned product manager, here is what I have learned over the years.

When Dashboards DON’T Add Value

Whether it’s B2C or enterprise software, we have all seen product fails in dashboards that:

When Dashboards DO Add Value

Product dashboards drive conversions and retention when:

Why do a Dashboard? Let’s Talk About the Data

Spend any time poring over analytics data, A/B data, status logs, or virtually any dataset and you will recognize a core truth: Data tends to look messy. Telemetry and log data may show gaps when systems went down. Most time series show some form of periodicity.

Very often outside factors affect time series.  For example, if a customer decides to scale back use of a product, the change in activity will show a dive in the data. Some data exhibits such strong variability that without complex smoothing it may tell no story.  Really good data displays handle variability honestly. They give context when data is missing or erratic.

The messiness of data hurts sales and renewal focused dashboards: Presales often goes to lengths to alter the underlying data in order to present a happy graph.

Collect the Right Data

What you show is only as good as what you collect. Many dashboards aimed at sales or renewals attempt to show Return on Investment (ROI). ROI is the black hole of death for dashboards. Why? Each customer makes very specific assumptions about missing or vague data-points and will ask about the methods used to collect and calculate any financial measure.

For example, your application tracks pricing across stores, and you collect prices, but not quantity, units per package, bundles, or line attributes like color. That makes any conclusions or displays suspect (retailers rarely discount black, if you are wondering). Making reports credible may require data you have no ability to collect.  ROI is important, but for complex products, it is hard to capture in a single visual and needs context.  

Make it fast…

Most of the time, the majority of reports or charts in your dashboard contain data which users do not drill into, and do not update frequently… yet most of the 3rd party tools folks use to make dashboards attempt independent live queries for each little visual. The result? Slooooooooowwwwww… loads.   Get slow enough and no one visits. A better strategy would pre cache and potentially pre-render charts. In my experience many reports really can work well with only daily or even weekly updates– and they load fast.

Present a Coherent Story about Time

Know that most all data used for dashboards tell some form of story about time: before and after, range in time, or a trend. When making a personal dash for executives and others, you see lots of time comparisons: To the previous quarter, the previous year, the previous week, the previous cohort. Even distributions (histograms) work within a time range.  Time is also the hardest part to build for—often your data is not structured in a way that easily allows “off the shelf” controls to display meaningful trends.

Be careful about mixing timescales—it can work very effectively if showing a form of cohort analysisCohort analysis is an analytical approach that breaks user data into related groups for analysis. Using cohorts allows you to focus in on different user behavior among groups such as conversions, churn rate, and more. but can also confuse readers.

No Scores Please!

Many product marketing people like the idea of a branded score or rating: 97 on a scale of 100. Scores may work well for consumer products, but they invite tragedy with enterprise users. In fact, enterprise users will call out any data irregularities. Anyone who spends anytime in a specialist role will pick apart scores very, very quickly.  Beware of single metrics that ruin the credibility of your dashboard.  

Data Never Explains Itself

Carefully chosen data? Excellent. Task focused views? Wonderful. Good annotation? Great. Just remember one thing:  Your dashboard is still a child of the market your product serves.  First-time users of a product must learn the lingo, purpose, and data rhythms of that market, whether it is digital advertising, credit cards, or security.  Design a product with that in mind.

Task-focused dashboards work because daily users take time to learn the lingo, rhythms, and interaction. Sales-focused dashboards tend to show superficial, pretty displays. You know a sales dashboard when you see world maps and meaningless, generic ‘arrows of progress.’  You know a task-focused dashboard when you see a dashboard that gets a real job done.

Design Dashboards for People Who Actually Log In

Active users for most products look to do some transaction or move to their main work space. If their daily work uses analytics, design will be straightforward. For example, cloud services may provide usage and billing analysis or status monitors. These kinds of tasks fit data dashboards: they deal well with messy data, can get actionable and atomic, and serve a clear task with a user you can get solid feedback from. As always, do you user research, so you know who you design for.

Successful dashboards focus on active users doing well-defined tasks, not some generic idea of insights. It is actually very difficult to manually sift thru aggregates in a data cube for salient features using typical controls. Doing that kind of manual sifting often requires so many different dimensions that datasets become painful to build and slow to use.

Clearly Separate Filters from Displays

Frequently, dashboards allow tapping on any visual to filter. Click on a bar in a bar chart and all the other grids, pies, big numbers filter out. This kind of feature goes against core principles of human-computer interaction:

 

Imagine the frustration of clicking on some random pie in a chart to see half your dashboard dim or disappear without any indication of what filtered what. Imagine users doing that in countless user study videos. Visual filters are powerful tools.  Remember to always do a few things for usability:

 

Don’t Fear the Grid

Dashboards can cram lots of sparklines, big numbers, and other heady visuals.  These kinds of dashboards are great for showing trends, relationships, weight. They show outliers without breaking a sweat.

However, many users just need an efficient grid of real numbers. Most of the personal dashboards I’ve done for executives or teams show tabular data.  Single, big numbers work very well as visual anchors but require context and social agreement on meaning.  Data doesn’t explain itself.  Don’t be afraid to design a flexible grid or tabular layout for your users.  It’s not pretty, but it works. 

Make it Easy for Users to Pull Data Out

Support the core user persona’s need to tell the product’s story to the budget owner.  When renewal time comes, the user that champions your product will need to justify the renewal with data. It amazes me when product managers hope that dozens of superficial charts or graphs from a dashboard that tells a fragile story ever reach the budget owner. Help your user personas making the business case!  Give them data in a form they can use with ease: easy-to-copy charts and data exports to spreadsheets. Help professional services answer “how did we do?” or “what were we working on?” Context is not always easy to productize.  Empower your core user personas and product champions the power to tell the product story with raw data.

Facilitate a Guided Conversation about Data

I love reading The Upshot in The New York Times. They do great work telling stories with data. Well designed renewal and status reports always tell good stories. Enterprise products often fill specialist needs and need explanation. Professional Services often crafts that story. Think about how to design data tools that create shareable data stories. These would honor the custody needs of sensitive data. Understand that professional services act as guide and personal contact when they jump on a screen-share or call to talk thru reports. We in technology discount the value of a personal touch. That kind of personal contact drives real connection in enterprise sales.

The need to protect data makes some common types of web sharing tricky: It is easy for Professional Services to make a clear separation of custody when they email reports to a customer. Using links to do the same thing often limits reach due to the need to control unauthorized access behind a login or signup.

Resources

No blog post can cover everything you need to know about dashboards.   Here are a couple of good books and websites for you to learn more about designing dashboards:

• Envisioning Information by Edward Tufte
• The Design of Everything Things by Don Norman
• Apple Design Guidelines (1995)
• The Upshot
• D3 

 

Takeaways