No-CodeFilterBuilder
Three years of complaints, fixed in two months. 92% faster task completion.
Asset Groups on Cisco's CX Cloud platform had been generating customer complaints for three years. Organizational reprioritization kept it unshipped despite existing designs. I picked it up as a stretch goal on a different module from my usual work, under a tight deadline.
I treated inherited designs and decisions as hypotheses, explored alternatives to stress-test them, scrapped what didn't hold up, and made the case to drop waterfall for agile delivery. Usability scores went from 3 to 6/7, time on task dropped 92%, and the MVP shipped in just 2 instead of 7 months.
Three years of complaints, zero fixes
Asset Groups had been generating complaints for three years. Improvements sat behind higher-priority platform work, and by the time designs were implementation-ready, the project had stalled with no design resource to carry it through implementation.
I had asked my manager for a stretch goal and was handed the project: a different module, inherited designs, a tight deadline. Rather than accept the designs at face value, I ran a quick in-house moderated test with colleagues to pressure-test them. The results were bad enough to convince the Product Owner to give me two weeks to fix the highest-impact failures before a single line of code was written.
Generic errors left users hanging
Every failure state surfaced the same generic string: network drop, empty dataset, misconfigured filter. Users had no context, no next step, and no way to tell what had actually gone wrong.
I worked with backend engineers to map every failure mode, then worked with a content designer to write a specific message and recovery action for each. More upfront coordination, but users could now self-recover without hunting for the cause.
No-code over SQL
I prototyped both a no-code filter builder and a full SQL editor rather than accepting inherited design decisions at face value. The PO assumed the technically advanced audience would demand a query language. Requirements analysis told a different story: only 1% of use cases were complex enough to need raw SQL.
We optimized for the 99% and built the no-code UI. With guardrails enforcing valid logic, it worked for experts and enabled junior admins who couldn't write queries at all. The 1% of edge cases we couldn't cover was a trade-off worth making.
Shipping 3x faster than planned
The project was scoped as waterfall: kick-off month 0, engineering planning month 1, release month 7. The team would ship when every feature, no matter how useful, was completely done.
I made the case for an agile approach and led a workshop with the engineering team. Scoping the MVP meant moving less critical features further out into the roadmap. That trade-off cut the timeline from 7 months to 2.
Impact
We shipped a 92% time-on-task improvement in 3x less time than estimated.
Remote unmoderated testing post-launch showed median usability scores improved from 3 to 6/7. Task completion dropped 92%. The MVP shipped in 2 months against a 7-month waterfall estimate.
Post-launch customer feedback through support tickets and community forums surfaced a gap for power users needing more expressive filters. This fed directly into a sequenced roadmap: v2.1 and/or connectors, v2.2 expanded operators (contains, starts with, ends with), v2.3 filter groups, v2.4 and IP-range filtering.
Next Project
MOSAIC Pivoting a 30-year-old Company
Turned siloed GIS tools into a unified city infrastructure SaaS. Majority of large customers upgraded.