Case study · Bondata
Bondata is an employee-engagement analytics platform for HR leads and managers in Finland. It turns pulse surveys into results across ~40 views and 25 chart types, so teams can see where engagement drops and set objectives to fix it.
Raise the share of managers who set an objective within 14 days of a survey from 18% to 30%, within two survey rounds after launch. HR leads renew Bondata when their managers act on the results, so this was the number that mattered to the client.
Tracked alongside it
A filtered result people can find and trust: task success and time on the baseline task
An objective one step away from the result behind it: task success on the baseline task
Fewer support tickets about filters, in equal periods before and after launch
Overall usability: SUS on the same test
Managers setting an objective within 14 days of a survey:
18%to31% (+13 pp)
Task success, creating an objective from a result:
3 of 8to8 of 8
Task success, finding a filtered result:
4 of 8to7 of 8
Filter-related support tickets, three months before and after launch:
33to12 (−64%)
Median time to find a filtered result:
74 sto22 s (−70%)
SUS score:
54to79 (+25 points)
Companies buy Bondata so that line managers act on their team's survey results, but only 18% set an objective after a survey closed. The results were hard to reach: filters were hidden in a collapsed side panel and did nothing until you pressed Refresh, and cards hid the one number that mattered. The goal was to raise that share to 30% within two survey rounds.
I planned and ran the research: the support tickets and usage analytics first, then interviews and a baseline test on the old product. Another designer took notes in every session.
214
support tickets reviewed
last six months, tagged by theme
~40
views in usage analytics
plus 25 chart types
8
baseline test participants
5 timed tasks and SUS
Also
Interviews with 4 stakeholders: the product owner, the founder, support and sales
In-depth interviews with 10 users from 5 client companies, 4 HR leads and 6 line managers, recruited through the client's customer success team
Managers open results once per survey, with about 15 minutes to answer one question: where did my team drop?
Visits that went to the dashboard and the detailed results report
Support tickets about filters
Most of them said “the numbers didn’t change”
Baseline participants who read a number from the previous filter setting
Without noticing: a confident wrong answer in an analytics product
Managers didn’t set objectives because the objective page was disconnected from the result behind it: “I’d have to remember which question it was”
Two-day workshop I ran with the client's product owner, the founder and the dev team
Review of filtering patterns in comparable analytics products
Effort vs. Impact matrix to prioritise the 36 ideas from the workshop
Moving filters above the content and applying them instantly
Putting the survey result on the objective card, and a "Set objective" action on every results card
A results summary a manager can scan in the 15 minutes before a team meeting, instead of reading it
Marking AI summaries before they are read
I tested Figma prototypes with 8 people (3 HR leads, 5 managers) on the baseline's five tasks. Filters need real data, so I retested them on staging with a client's survey, again with 8 people. Task success and time in the results come from that round.
I set the visual language and rebuilt the design system as a Material UI theme: 44 component sets, 136 colour variables, light and dark. Two designers applied it across part of the ~40 views under my review. Structure changed only where 82% of visits go: the dashboard and the results report. At twice-weekly client reviews I defended decisions with test results.
Decision 01
Before: find a collapsed side panel, open it, set up to 20 filters, press Refresh. Now filters sit in a row above the content and apply instantly. On scroll the row collapses into a sticky button; extra filters fold into "More".
Decision 02
The old card had a title, description and status, on a separate page. Now an objective carries its result: question, group, index, score and comments. "Set objective" on every results card turns a weak result into an objective without leaving the page.
Decision 03
The card was plain text at one size, benchmarks in brackets at line ends. Now response rate and comments are two figures on top, each strength and development area is a row led by its percentage with a green or red bar, and the benchmark is a tag. Downloads became links.
Decision 04
Three open paragraphs ended with "AI-generated summary may contain errors": you learned the source after reading. Now an "AI Summary" badge sits on top, the card wears the brand gradient no chart uses, and the text collapses to two lines behind "Read more".
Managers setting an objective within 14 days of a survey:
18%to31% (+13 pp)
Task success, creating an objective from a result:
3 of 8to8 of 8
Task success, finding a filtered result:
4 of 8to7 of 8
Filter-related support tickets, three months before and after launch:
33to12 (−64%)
Median time to find a filtered result:
74 sto22 s (−70%)
SUS score:
54to79 (+25 points)
Evoke
Art direction for an AI companion that must not behave like one
(view)
Dishcover
A culinary app with easy, step-by-step recipes for every taste
(view)
Your Vet
In progress
