Tasks, the activity log of a private cloud’s control plane
And by letting admins scan or dig in without leaving the tableAnd by letting admins scan or dig in without leaving the tablePrism Central is the console Nutanix customers use to run their private cloud. Every action in it, from creating a VM to upgrading a cluster, becomes a task, and the Tasks screen is where IT teams check on those tasks. Customers and SREs were complaining about it, and when we surveyed customers, 52% said it needed to improve.
I led the research and the design. Navdeep Bagga was my collaborator and guide: he helped plan the research, brainstormed with me, and critiqued the designs. Rachel Chang was the PM, Monica Jeyachandran the engineering manager, and Romulo Rosinha built the frontend.
The redesign shipped. It scored 86.7 on the System Usability Scale in testing, and a year on, customer issues about Tasks are almost nil. Its patterns now guide the design of Alerts, Events and Audits.
The Tasks screen is like a TV remote. As long as it works, nobody notices it. Once it breaks, people start cussing, and people were cussing at Tasks.
Nobody visits it for its own sake. A cloud admin lands there on a Monday morning with one question: what happened to the host on Friday around 10 PM? An engineer checks whether a VM finished creating so they can deploy. An SRE on a customer call needs to know if an upgrade is stuck.
The screen made even those questions hard. It was 321 tasks in one flat table, and finding the one you cared about meant paging through noise.
Tasks had no PM driving it. Navdeep and I had to evangelize the problem ourselves to get it on the radar. Two artifacts did it: the heuristic evaluation won us a PM, and the storyboard won engineering's backing.
Customers had logged around 147 issues against Tasks, so we knew it was a problem. The research was to find out where.
Six findings shaped everything after:
We would know it was fixed by two measures: in the short term, how the redesign did in usability tests, and after launch, the trend in Jira tickets about Tasks.
I turned the research into a storyboard, "Finding the weekend memory snatcher." It follows Kirk, a cloud admin playing detective on a Monday morning, through his frustration with the old screen, and shows where a redesign could remove it.
For the table itself I studied vSphere's task UI, which users and stakeholders kept pointing to as the benchmark, then explored three layouts:
C won because it served all three jobs: scanning for status, digging into a failure, and following how tasks relate.
One thing we got wrong: logs. We believed users would care about them, so we tried surfacing logs on the screen for failed tasks. They did not, and we cut it.
After a few more rounds of visual and interaction polish, the redesign shipped.
Did it fix the problem? By both of the measures we set, yes.
In testing. I ran seven validation sessions with customers from Xtribe. The redesign scored 86.7 on the System Usability Scale, where 68 is the usual average.
"This is a billion times better!" "It's party time!" "Less jumping around." "The Filters are absolutely amazing!"
After launch. The count of customer issues, around 147 before, is almost nil a year on, and the complaints that do come in are feature requests.
Beyond the screen. Design driving this had a side effect: we were able to standardize what a task in Prism Central is. Service teams had to adopt that definition, and send what belongs in Events or Audits there instead, which was our answer to the three blurring together. The patterns and decisions from the redesign now guide the design of Alerts, Events and Audits too.
Get buy-in with richer artifacts. A list of 21 issues won us a PM, and a storyboard won us engineering. The storyboard compressed weeks of research into something a stakeholder could take in within minutes, and I have used artifacts like it on every project since.